Capability, pricing and access changes, and notable events. Each entry shows its date and sources. We only use the word "breaking" when the primary source's own date is within 48 hours. Unconfirmed reports are labelled with a confidence level and are never presented as fact.
2026-11-09 · Buyer accuracy · Hedwig
Current official pages now document Free with 100 one-time AI credits and one email account, Pro and Team monthly/annual prices, Mac/Windows downloads and Outlook alongside Gmail. We explain the shared AI credit meter, email Outbox approval and the distinction between protected user content and advertising identifiers. Documentation review only; no hands-on trial or measured task cost.
Confidence: Primary sources (vendor pricing, documentation and privacy policy) · Sources: Hedwig pricingCredits and dataPrivacy policy
2026-11-05 · Buyer questions · Replit and Copilot Studio
Replit: Slack account linking, thread ownership and channel-visible replies are now explained separately from Pro/Enterprise existing-repo Cloud Agents. Copilot Studio: maker consumption reports and preview admin cost controls are documented without claiming universal hard caps or automatic PAYG. Documentation review only; no new integration tests.
Confidence: Primary sources (vendor documentation and dated announcements) · Sources: Replit Slack docsReplit existing-repo announcementMicrosoft October update, published 4 November
2026-11-05 · Pricing update · Kiro
Team and Enterprise credit pooling costs $20 per seat per month, with a two-seat minimum ($40/month before tax for two seats contributing 1,000 shared monthly credits). Each seat contributes 500 credits, not the 1,000 in the legacy Enterprise Pro tier. Included credits expire monthly; purchased add-ons cost $0.04 each and expire after 12 months. Pooling uses included credits first, then the add-ons nearest expiry, and does not use overages. Pooling currently requires a Kiro profile using AWS IAM Identity Center; profiles connected directly to an external identity provider cannot use it. New Enterprise customers use pooling; existing subscribers can stay on legacy tiers or opt in. Switching subscribers to sharing is irreversible. If cancellation of a legacy subscription fails during migration, it can remain active and billed. Check the purchase summary: the AWS account determines prepaid or postpaid billing.
Confidence: Primary source (vendor pricing and documentation) · Sources: Kiro announcement dated 4 NovemberKiro pooling documentation
2026-11-04 · Agent update · Agentforce
Salesforce's 4 November 2026 announcement lists Google Workspace and Microsoft 365 connectivity, Conversation History, Skills and the $50/user/month SKU as available now. Teams, iOS, Mac/Windows apps, the MCP server and a Flex Credits customer trial are listed for general availability during November, not all confirmed live today. Daily Digest and scheduled tasks are expected in December. Availability may vary by region and agreement; the $50 price does not establish unlimited use of every Agentforce feature. Our pricing review also found that retrieved text lists $2/conversation while the live browser card displays no amount. The guide and profile now disclose this disagreement rather than present a universal price.
Confidence: Primary source (vendor announcement and pricing page) · Sources: salesforce.com announcementsalesforce.com pricing
2026-11-04 · Agent update · Agentforce
Salesforce's updated Flex Credits Rate Card (dated 3 November 2026) adds IT and HR Employee Agent Resolutions as a usage type at 400 Flex Credits per resolution (0 in sandbox), the same rate as Help Agent Resolutions. All other multipliers we track are unchanged: standard actions 20, voice actions 30, prompts 2/2/4/16, speech and translation rates, Personalization 100 per 1,000 decisions, and the no-rollover rule. The per-conversation price remains unpublished.
Confidence: Primary source (vendor rate card) · Sources: salesforce.com
2026-11-03 · Update · 1 agent
The Copilot Studio profile now carries Microsoft's per-feature Copilot Credit rates (Microsoft Learn, read 3 November 2026): classic answer 1 credit, generative answer 2, agent action 5, tenant graph grounding 10, agent flow actions 13 per 100, content processing 8 per page. Microsoft's stacked example - graph grounding plus a generative answer - totals 12 credits per complex prompt, about $0.12 at the pay-as-you-go rate; a 25,000-credit pack covers about 2,083 such answers arithmetically. This answers the recurring community question about the cost of knowledge-base-heavy agents.
Confidence: Vendor documentation read 3 Nov 2026 · Sources: Microsoft Learn: billing rates and management
2026-11-03 · Update · 2 agents
Copilot Studio replenishment timing; Manus Cue billing clarified
Two documentation clarifications from community-driven questions. Copilot Studio: prepaid capacity-pack credits replenish at the start of every monthly billing period (the subscription's billing cycle, not a fixed calendar date), and without a pay-as-you-go safety net, exhausted prepaid credits suspend Copilot Chat until replenishment (Microsoft Learn, read 3 November 2026). Manus Cue: the phone-number subscription is separate from the Manus membership and phone usage draws on a weekly Cue allowance, not Manus credits; the subscription price remains unpublished (Manus help, read 3 November 2026).
Confidence: Vendor documentation read 3 Nov 2026 · Sources: Microsoft Learn: capacity packsManus help: membership pricing
2026-11-03 · Addition · 1 comparison
New comparison: /compare/lovable-vs-replit-agent/ puts our hands-on Replit Agent pilot (2 November 2026: tip calculator with PostgreSQL history, three sequential prompts, all checks passed) next to Lovable's documented pricing (read 2 November 2026: free daily build grants, Pro from $25/month for 100 credits). No winner claimed: both bill in credits that mean different things, and Lovable is documentation-only because we could not register to test it.
Confidence: Replit pilot 2 Nov 2026; Lovable docs read 2 Nov 2026 · Sources: Lovable docs: subscription plans
2026-11-03 · Addition · 1 comparison
New comparison: /compare/bolt-vs-lovable/ puts our hands-on Bolt pilot (three build-and-edit pairs on the free plan, 22-24 October 2026) next to Lovable's current documented pricing (read 2 November 2026: free daily build grants, Pro from $25/month for 100 credits, annual at 10x monthly). Bolt's pricing was re-read the same day (Pro $25/month billed monthly). No winner claimed: tokens and credits are not convertible, and Lovable is documentation-only because we could not register to test it.
Confidence: Bolt pilot 22-24 Oct 2026; both vendors' pricing read 2 Nov 2026 · Sources: bolt.new/pricingLovable docs: subscription plans
2026-11-02 · Update · 1 agent
Lovable's paid-plan prices, previously not captured, are now documented from the vendor's subscription-plans page (read 2 November 2026): Pro from $25/month for 100 monthly credits, Business from $50/month for 100 credits, tiers up to 10,000 credits/month on both, annual billing at 10x the monthly price. Free plan unchanged: 5 build credits a day (up to 30 a month) plus monthly Cloud and AI grants.
Confidence: Vendor documentation read 2 Nov 2026 · Sources: Lovable docs: subscription plans
2026-11-02 · Addition · 1 guide
New guide: /guides/bolt-pricing-explained/ explains Bolt's token pricing (Free $0: 1M tokens/month with a 300K daily cap; Pro $25/month billed monthly with 10M tokens and rollover; Teams $30/member/month) as read on bolt.new/pricing on 2 November 2026, and adds our hands-on pilot readings: one build-and-edit pair consumed roughly 155K-195K tokens, about 52-65% of the free daily cap. The 22 October annual-billing figures ($18/$27) are labelled historical: no annual option was visible in the 2 November read.
Confidence: Vendor pricing read 2 Nov 2026; pilot readings 22-24 Oct 2026 · Sources: bolt.new/pricing
2026-11-01 · Addition · 1 agent
78 agents: DeepSeek dsh-superagent added
DeepSeek's Super-Agent, covered widely on 1 November, is documented from the vendor's own repository rather than the press framing: it is an MIT-licensed research prototype of a proactive coding agent for one GitHub repository, built as a bundle for DeepSeek Harness, and the README states plainly that it is not an official DeepSeek product or production-ready software. It listens to GitHub (optionally Feishu), acts only when an allow-listed person asks or a team-written rule allows, and answers in the original conversation, typically with a pull request. Code is free; it needs DeepSeek Harness with a model provider you configure.
Confidence: vendor_docs · Sources: GitHub repository
2026-11-01 · Re-check · Kiro
Kiro pricing guide re-verified: upgrade offer changed to a free month plus 500 trial credits
The /guides/kiro-pricing-explained/ guide (first published 26 October) was re-read against Kiro's pricing page and FAQ on 1 Nov 2026. The plan line-up is unchanged (Free 50 credits; Pro $20; Pro+ $40; Pro Max $100; Power $200 per user/month). What changed is the first-upgrade offer: a free rest-of-month plus 500 trial credits (usable within 12 months) replaces the previous $20 prorated first-upgrade credit; Kiro says users who upgraded before 1 November 2026 keep the $20 discount and unused credits. Kiro's proration example (533 credits for a half-month at Pro) implies roughly 1,000 monthly credits on Pro, though per-tier allowances remain unpublished. Monthly plan credits still do not roll over; add-on credits remain $0.04 each with 12-month expiry.
Confidence: high · Sources: kiro.dev
2026-11-01 · Launch watch
DeepSeek open-sources a coding agent prototype; Microsoft releases a small computer-use agent
DeepSeek's Harness team published dsh-superagent-prototype on GitHub (31 Oct), an open-source router-worker coding agent positioned against Claude Code; it is a prototype repository, not a priced product, so no directory profile yet. Microsoft published Mage-VL-Cowork, a 4B-parameter hybrid GUI+code computer-use agent with a tech report and Hugging Face weights - a research release, not a sign-up product. Both are tracked for directory coverage if they become products with published terms. Sources: github.com/deepseek-ai/dsh-superagent-prototype, microsoft.github.io/Mage/cowork, huggingface.co/microsoft/Mage-VL-Cowork (all read 1 Nov 2026).
Confidence: high · Sources: github.commicrosoft.github.iohuggingface.co
2026-10-31 · Re-check · Claude Code
Claude plans renamed: Max 5x and Ultra 10x weekly replace Max5/Max20; monthly API credits documented
Anthropic's pricing page now lists four individual plans: Free, Pro (1x, includes Claude Code), Max (5x Pro on 5-hour and weekly windows, $100/month Claude Platform API credits) and Ultra (no 5-hour limit for verified accounts, 10x Pro weekly, $200/month API credits). GBP view: Pro GBP15/mo annual, Max GBP90/mo, Ultra GBP180/mo incl. VAT. Team seats documented at 1.25x (Standard) and 6.25x (Premium) of Pro per session with weekly limits and per-member metering; weekly multiples are published for individual plans (1x/5x/10x) but not for Team seats. Limit resets exist but are only given occasionally, not sold on demand.
Confidence: high · Sources: claude.comsupport.claude.comsupport.claude.comsupport.claude.com
2026-10-30 · Re-check · Copilot Studio
Copilot Studio pricing re-verified: $200/pack/month for 25,000 Copilot Credits unchanged; harness framing documented
Full-record re-check against Microsoft's pricing page and FAQ: the tenant-wide licence at $200/pack/month for 25,000 Copilot Credits, the pay-as-you-go meter (Azure subscription required) and the pre-purchase plan with up to 20% savings are all unchanged. Microsoft now frames rights by harness - Microsoft 365 Copilot includes Copilot Chat and Standard harnesses for internal agents at no additional cost, while the standalone licence adds the GitHub Copilot harness and external-channel publishing. Microsoft also lists a free trial, an in-context Agent Builder inside Microsoft Copilot, and a Copilot Studio Lite footnote on some offers.
Confidence: high · Sources: microsoft.commicrosoft.com
2026-10-28 · Update · Muse
Muse pricing documented: free weekly limit, Power $20/mo, Maximum $100/mo
Meta's subscription help pages now document Muse's plans: free with a weekly usage limit, Power at $20/month (500M tokens/week) and Maximum at $100/month (3B tokens/week) - previously unpublished. A full re-check of nine Meta help pages also established the connector model (retrieve-only per connector, custom connectors unreviewed), the granular approval system, the built-in browser with a Secure Credentials Store, and the training opt-out that starts on. Regional availability is still unpublished: Japan (23 Oct) and Singapore (28 Oct) rollouts are press-reported, not vendor-listed.
Confidence: vendor_docs · Sources: meta.commeta.com
2026-10-27 · Update · Fo by Wajo
Fo by Wajo: free plans documented again, MCP server launched
Wajo's help FAQ now documents Fo mini (free, daily limit) and OG Fo (free, unlimited tasks, invite-gated), with Enterprise by quote and Fo Pro 'coming soon' - still no public price list. Separately, Wajo launched a hosted MCP server: Claude, Cursor, ChatGPT and VS Code can build and drive Wajo agents, and a /fo profile lets consumer apps host Fo. The vendor's 19 October launch posts also show physical-device control (lamps, vacuums) over webhooks and MCP; not yet in vendor documentation and not tested.
Confidence: vendor_docs · Sources: wajo.aiwajo.aiwajo.ai
2026-10-26 · Guide · Kiro
New guide: Kiro pricing explained
Kiro’s five tiers, the credit model, add-on credits at $0.04, the no-rollover rule and the unpublished tier allowances - all from Kiro’s pricing page and billing FAQ, checked 26 October. GSC-evidenced demand (kiro pricing queries, 28d to 23 Oct). Kiro’s profile pricing record re-verified the same day.
Confidence: vendor_docs · Sources: kiro.devkiro.dev
2026-10-26 · Addition · 1 agent
77 agents: Gemini Enterprise added
Google’s universal agent for work, announced at Gemini at Work on 8 October: one prompt box for knowledge work, media and code, per-job model orchestration across Gemini and Claude, and enterprise governance. Pricing verified on the Business edition page (from $21/seat/month, 30-day trial); the Service Specific Terms training restriction and zero-data-retention caveats verified in Google’s documentation.
Confidence: vendor_docs · Sources: cloud.google.comcloud.google.comdocs.cloud.google.com
2026-10-26 · Addition · 1 agent
76 agents: Eney (MacPaw) added
Eney is MacPaw’s proactive assistant for Mac and iPhone, launched 15 October: on-device context across apps and files, confirm-first actions, meeting transcription, Eney Wand and ready-made workflows. Pricing, permission model and the privacy notice’s local-model training clause verified on MacPaw’s pages: 14-day trial, then Setapp AI-credit subscriptions from $14.99/month.
Confidence: vendor_docs · Sources: macpaw.commacpaw.commacpaw.com
2026-10-26 · Additions · 2 agents
75 agents: Claude Duet and Sundae added
Two vendor-documented launches from last week. Claude Duet is Anthropic’s proactive agent (US research preview, waitlist, included with Pro/Max) - pricing, approval-first action design, one-time purchase cards and hard off-limits actions verified on Anthropic’s product, safety and help pages. Sundae is sundae_bar’s text-message personal agent (open beta, iMessage/RCS, free daily limit) with the Plus plan price still unpublished; permissions and privacy verified on its launch site and privacy policy.
Confidence: vendor_docs · Sources: claude.comclaude.comsupport.claude.comclaude.comsundae.mdsundaebar.aisundaebar.aisundaebar.ai
2026-10-25 · Correction · Manus
Manus was listed under errands, travel and food. Its own homepage and documentation centre on producing digital deliverables - slides, websites, games and files - so it now sits with work and business agents under the same vendor-positioning rule applied to Genspark. Its newer Cue app remains listed separately as a personal agent.
Confidence: Vendor site positioning (manus.im, read 25 October 2026) · Sources: manus.im
2026-10-25 · Correction · Genspark
Genspark was listed under errands, travel and food. Its own positioning is an all-in-one AI workspace producing documents, slides, sheets, dashboards and code, so it now sits with work and business agents. Pricing and capability records are unchanged.
Confidence: Vendor site positioning (genspark.ai, read 25 October 2026) · Sources: genspark.ai
2026-10-25 · New agent · Skye
Signull Labs’ Skye, previously invitation-only, is now live on the App Store (free, no in-app purchases listed, iOS 26+). It connects calendar, Gmail, spending via Plaid, Apple Health, Whoop, Slack and location into one ranked home-screen feed, with confirm-first agentic actions per its terms.
Confidence: Primary sources (vendor site, App Store listing, privacy policy, terms) · Sources: apps.apple.comskyeapp.ai
2026-10-25 · New agent records
Windsurf (Cognition AI): agentic editor with Tab, inline edits, frontier models and Devin Cloud agents - Free $0, Pro $20/month, Max $200/month, Team $80/month plus $40 per full dev seat. Its platform terms are the privacy story: customer data may train models by default, and only paid tiers can opt out, which then enables zero data retention with model providers. Warp (Denver Technologies): agentic terminal with Agent CLI, cloud agents and model choice - Free $0, Build $20/month with $20 of usage included, Max $200/month, Business $50/user/month; add-on usage rolls over month to month. Warp's terms let it train on licensee data, with data controls configurable per plan.
Confidence: High - vendor pricing pages, privacy policies and terms read directly on 25 October 2026; vendor statements, not independently tested. · Sources: Windsurf pricingCognition platform termsCognition privacyWarp pricingWarp termsWarp privacy
2026-10-25 · New agent records
Two long-queued records land. Stanley (Find Community, dba Stan) is an AI head of content for creators on iMessage, Telegram and web: it connects Instagram, LinkedIn and YouTube, drafts scripts and captions, and schedules posts - free 3-day trial, then $44.99/month per the App Store (its own site publishes no prices). Genspark (MainFunc) is an all-in-one AI workspace whose Super Agent produces documents, slides, sheets and live dashboards: Free $0 (100 credits a day), Plus $24.99/month, Pro $249.99/month, Team $40/seat/month per the vendor help centre. Both privacy policies reviewed.
Confidence: High - vendor pages, App Store listing and help centre read directly on 25 October 2026; vendor statements, not independently tested. · Sources: StanleyStanley App StoreStan privacy policyGensparkGenspark membership plansGenspark privacy
2026-10-24 · Agent records updated
Four records re-verified. Del's terms now include a suicide and self-harm safety protocol (US 988 referrals; it does not contact emergency services) and its privacy policy discloses limited human review of requests, with in-conversation disclosure. Brea re-issued its privacy policy on 23 October 2026: brea.ai runs Meta and OpenAI advertising pixels that receive hashed email, name and billing location on signup or subscription; saved site logins are encrypted; sign-in is Apple or Google. Its 'Free, with every feature included' offer stands. TinyNature (beta, no pricing, unverified SOC 2 badge) and HireAlpha ($19 per persona per month, waitlist-first) are unchanged.
Confidence: Vendor pages read 24 October 2026; vendor statements, not independently tested. · Sources: Del termsDel privacyBrea privacy (23 Oct 2026)Brea homepagetinyNature homepageHireAlpha homepage
2026-10-24 · Agent records updated
Zapia now publishes its plans on the homepage: Free $0 (up to 20 scheduled tasks), Pro $9.99/month and Ultra $99.99/month, both with 7-day trials. It has repositioned as 'the AI that runs your WhatsApp' for Latin America, with connectors for WhatsApp, Gmail, Google Calendar, Drive, Outlook, Slack and iFood. Its privacy policy (5 October 2026) names the third-party AI providers that process user content and discloses affiliate compensation on product recommendations. Grok Bot (verified in a live browser) keeps its $30/$100/$200 monthly tiers with weekly usage and token-cost overage, is now included with Teams and Enterprise plans, and now downloads for Linux, macOS, Windows, iOS and Android; new documented features include routines and multi-bot threads on one shared cloud computer per user. Asmi was re-verified with no changes (first task free, Pro $10, Ultra $49 per month).
Confidence: High - official pages read directly on 24 October 2026; Grok Bot verified in a live browser · Sources: ZapiaZapia privacy policyGrok BotAsmi pricing
2026-10-24 · Agent records updated
Tomo's FAQ now publishes a starting price: the base plan is from $19.99/month with pricier plans for higher messaging limits, cancel anytime in the dashboard - though pricing may vary by user (promotions). Tomo's privacy policy also gained an Apple Health section (health data powers the agent; it is never sold, used for ads or model training) and a CCPA Do-Not-Share opt-out. Orchid's setup number changed to +1 (415) 605-3216, and the company launched Keiki, a platform for building customer-facing agents across iMessage, SMS, WhatsApp, Slack, Telegram and email; its $29.99/month no-free-tier pricing is unchanged. asaply was re-verified with no material changes (per-order fee, early access, legal page still dated 11 October 2026).
Confidence: High - official pages read directly on 24 October 2026 · Sources: Tomo about and FAQTomo privacy policyOrchidIntroducing Keikiasaply legal
2026-10-24 · Agent records updated
Shuffle (DNFT) has pivoted from a group-chat assistant to an AI social media manager for small businesses: it makes and publishes posts, answers DMs, comments and Google reviews, and you approve each post by text. Plans are now published: Starter $99, Business $199 and Agency $299 per month with a free 3-video start; the old chat product survives as a free companion with an $8/month Pro tier. Kicker (Companion Intelligence) has launched: the waitlist is gone, accounts are free with phone or Google sign-in, and it connects Oura, WHOOP and Fitbit - but paid plan prices are still unpublished and its privacy policy still describes the earlier voice-call platform. Boski was re-verified: its privacy policy still says user content is not used to train foundation models, its terms (effective 20 August 2026) bar using the service to benchmark a competing product, e-mail is now a listed channel, and pricing is still not published.
Confidence: High - official pages read directly on 24 October 2026 · Sources: Shuffle plansKicker FAQBoski privacy policyBoski terms
2026-10-24 · Guide
Flex Credits ($500 per 100,000, 20 per action, 30 per voice action), the Conversations model with no published price, employee add-ons ($125/user/month), Industries add-ons ($150), Max Edition (from $550 with 2.75M credits per org per year), the $5 User License and $50 Coworker licence, buying models and no-rollover rules. All from Salesforce's pricing page and both rate cards.
Confidence: High · Sources: Salesforce Agentforce pricingFlex Credits Rate Card
2026-10-24 · Pricing correction
Our Agentforce record previously showed $2 per conversation. That figure does not appear on Salesforce's current pricing page or rate cards (checked 24 October 2026), so the record now says the per-conversation model is listed without a published price. Flex Credits ($500 per 100,000), add-ons ($125/user/month) and Max Edition (from $550) are unchanged.
Confidence: High · Sources: Salesforce Agentforce pricing
2026-10-24 · New agent record
Asmi (Humint Labs) is a personal assistant you text on WhatsApp, SMS or iMessage; it calls, texts and emails businesses to complete tasks such as bookings, appointments, quotes and follow-ups. First task free, then $10/month (20 tasks) or $49/month (100 tasks). Calls on your behalf identify themselves as AI and disclose recording; marketing, debt-collection, emergency, political, harassment and legal-threat calls are refused. Record built from the vendor's homepage, pricing, privacy policy and terms, read 24 October 2026. No hands-on testing.
Confidence: High - homepage, pricing page, privacy policy and terms of service read directly on 24 October 2026 · Sources: AsmiAsmi pricingAsmi privacy policyAsmi terms of service
2026-10-22 · Agent records updated
All 67 agent records were re-checked against live vendor pages on 22 October 2026. Notable changes: Folk is now free with every feature included (100 messages a day; the $9.99 Pro tier is gone) and runs a $15 referral programme. Brea's homepage now says 'Free, with every feature included' and its pricing page is gone. Cursor replaced its $20 Individual plan with Free, Plus $8, Premium $30, Super $100 and Ultra $200 tiers. Bolt cut its annual prices (Pro $18/month billed annually, Teams $27). Grok Bot published pricing for the first time (Premium $30, Super $100 and Ultra $200 per month, billed monthly, with weekly usage included and token-cost overage). Mitra now lists three tiers from $30/month. Case no longer publishes its $29 price. Fin moved to fin.ai, Codex pricing moved to learn.chatgpt.com, Wabi is now 'Wabi: AI Assistant' on the App Store, and GitHub's Copilot cloud agent docs moved. Perplexity renamed its paid plans (Select $17/month and Max $167/month, both billed annually), and Manus confirmed paid plans from $20/month with a free daily-credit tier. Prices for Comma, Claude Code, Agentforce, Kiro, Pally, Replit, Jules, Underdog and 15 more agents were re-verified unchanged against their official pricing pages.
Confidence: High - live vendor pages read directly during the 22 October 2026 audit; Cursor and Grok Bot pricing verified in a live browser · Sources: Folk pricingBreaCursor pricingBolt pricingGrok BotMitra pricing
2026-10-22 · Agent record updated
Boba's maker humans& announced the launch on 22 October 2026. The product moved from getboba.ai to boba.app, keeps its free iMessage one-on-one and group-chat assistant, and adds Boba Web (free, no app, no sign-up). The privacy notice was refreshed on 21 October 2026; it still permits model training on messages, with stricter treatment for Google User Data.
Confidence: High - vendor pages and launch post read directly · Sources: Boba homepageBoba privacy noticeLaunch announcement on X
2026-10-22 · New agent records
Added five records. Del is a personal assistant you text for briefings, refunds, cancellations and monitoring (free two-week start, no paid price published). tinyNature runs on WhatsApp, iMessage and Telegram with email triage, automations and tracking (beta, no pricing published). HireAlpha puts three separate AI personas in iMessage, each with its own number and memory, at $19/month per persona from a waitlist. Zapia is a proactive assistant in its own app and web (free start advertised; Pro/Ultra prices not published) and Grok Bot is xAI's desktop/iOS AI teammate that works inside your logged-in tools (no public price); both are directory records, not text-message table entries. The text-message comparison now covers 30 assistants.
Confidence: High - vendor pages read directly · Sources: DeltinyNatureHireAlphaZapiaGrok Bot
2026-10-21 · New agent records
Added records for Orchid (personal assistant inside iMessage: Gmail triage, drafts in your voice, approved scheduling; paid from $29.99/mo with no free tier) and asaply (text-first errand runner for coffee, rides, tickets and group errands; per-order service fee, no subscription, early-access availability varies by city). Both join the text-message comparison, which now covers 27 assistants.
Confidence: High - vendor pages read directly · Sources: Orchidasaply
2026-10-21 · New agent records
Added records for Shuffle (AI teammate for iMessage and SMS group chats; free start claimed but no pricing or privacy page found), Kicker (Coming-Soon proactive assistant from Companion Intelligence that picks daily priorities and calls businesses; waitlist only) and Tomo (text-based personal AI from Mapo Labs for goals, reminders and life admin; terms describe paid auto-renewing subscriptions but no public prices). Shuffle and Tomo join the text-message assistants table; Kicker does not, because only outbound calling is documented, not two-way messaging. Documentation research only; nothing tested hands-on.
Confidence: Official homepages, about pages, terms and privacy policies · Sources: Shuffle homepageKicker homepageKicker privacy policyTomo aboutTomo privacy policyTomo terms
2026-10-20 · New agent records
Added records for Boba (assistant inside Apple Messages group chats, free per its homepage, privacy notice permits model training on messages), Boski (cross-channel assistant on iMessage, WhatsApp, X, Safari and app; privacy policy says no foundation-model training on user content; no public pricing) and Brea (hosted personal agents with integrations, schedules and browser automation; Free tier and $9/month Starter, Pro price not established). All three join the text-message assistants comparison table. Documentation research only; nothing tested hands-on.
Confidence: Official homepages, pricing pages and privacy policies · Sources: Boba homepageBoba privacy noticeBoski homepageBoski privacy policyBrea homepageBrea pricingBrea privacy policy
2026-10-16 · Documentation expansion
Added eight evidence rows and a narrow setup path to each agent. Separated Kiro surface/version controls and abuse retention, Lindy connection identity and shared-Slack guardrails, and Poke integration scope and Maximum Privacy. Unknowns and vendor-page conflicts remain visible. Documentation, not hands-on tests.
Confidence: Official product, pricing, permission and privacy documentation · Sources: Kiro current permissionsLindy integration scopePoke privacy notice
2026-10-16 · Pricing, access and privacy correction
Replaced old 3-app/$25 plan claims with the current free core allowance, paid Cloud Mac add-on and invitation-only VIP. Separated iMessage/RCS/Telegram chat from connected WhatsApp, and documented terms-based approval exceptions, model-training opt-out and retention limits. Vendor documentation, not a hands-on test.
Confidence: Official pricing, terms and privacy policy · Sources: Current pricing and channelsApproval exceptions and paid plansTraining, wallet approvals and retention
2026-10-14 · Pricing and control correction
Corrected Team pricing to $49.99/user/month for 5,000 credits. Integration docs scope write guardrails to shared Slack threads, while the pricing FAQ still makes a broader approval claim. The record now preserves that conflict rather than promising approval on every channel. Documentation, not a billing or security test.
Confidence: Official pricing and integration documentation · Sources: Current pricing and FAQGuardrail channel and setting scope
2026-10-14 · Documentation expansion
Added eight evidence rows and narrow setup steps for each. Fin distinguishes timeout-based email input and connector retention; Codex separates current and legacy cloud credentials; LangGraph explains checkpoint and replay prerequisites. Vendor documentation, not tests.
Confidence: Official documentation · Sources: Fin escalation and email inputCurrent Codex CloudLangGraph interrupts
2026-10-13 · Pricing correction
Corrected the outcome guide to distinguish $0.99 explicitly marked Self-serve routing from $9.99 full qualification. Existing routing outcomes are not automatically repriced; meeting and follow-up restrictions apply. Based on current Intercom documentation, not a billing test.
Confidence: Official outcome documentation · Sources: Intercom outcome definitions
2026-10-13 · Evidence expansion
Added eight documented feature rows and narrow setup checklists for Cursor, Jules and GitHub Copilot cloud agent. Cursor privacy-language mismatch, Jules automatic API plan approval and GitHub firewall/workflow limits stay explicit. GitHub variable credit allowances and conflicting Max FAQ wording are labelled. Documentation research, not hands-on tests.
Confidence: Primary documentation, not tested · Sources: Cursor securityJules APIGitHub controls
2026-10-12 · Evidence expansion
Added Agentforce licence and charging-model detail with a labelled vendor example, connected Copilot licence scope to its pricing guide, and added dated authentication/confirmation evidence to the governance guide. Fin and Copilot pricing guides now include a secondary further-reading reference. No purchases or hands-on tests.
Confidence: Primary documentation, not tested · Sources: Salesforce pricingCopilot licensing
2026-10-11 · Evidence rollout
Added 24 distinct source-linked rows and narrow setup. Clarified maker-auth triggers, Devin permissions and training, and Replit Free/Power/Max billing with paid-action confirmation. Replit billing guide and coding comparison updated together. No hands-on tests or rankings.
Confidence: Fresh official documentation, not tested · Sources: Copilot triggersDevin permissionsReplit billing
2026-10-10 · Product release
Published the directory data as a source-linked JSON list and per-agent records with schema metadata, dates and explicit evidence limits. No testing scores, rankings or paid product claims.
Confidence: Live data interface; documentation research, not agent tests · Sources: JSON list
2026-10-10 · Correction
Official pricing and billing pages describe different Teams base-charge models. The guide now labels this conflict rather than treating either as a confirmed quote, and corrects unlimited membership to a 200-user self-serve cap.
Confidence: Official sources disagree; not tested · Sources: Devin pricingBilling documentation
2026-10-10 · Evidence rollout
Added distinct source-linked tables and narrow-start steps. Cue wallet limits remain unknown; Folk calling and pricing sources disagree; workspace shared authentication, write approvals, API limits and plan-dependent costs are explained. Vendor documentation, not product tests.
Confidence: Official documentation, not tested · Sources: Cue documentationFolk privacyWorkspace agents
2026-10-09 · New comparison
Added a retailer-facing comparison of seven shopping agents and checkout products: payment routes, approvals, merchant controls and dated primary sources. No purchase tests or rankings.
Confidence: Primary-source documentation, not tested · Sources: Link agent paymentsMuse payment rulesShopify checkout
2026-10-09 · Evidence rollout
Re-read primary sources and added product-specific coverage tables and narrow-start steps. Corrected Claude Code Auto vs Manual starting-mode claims and related guides; expanded Fo spending/human/call limits, dots proactive research and memory controls, and Carly free/paid tools. No hands-on tests or automatic ranking claims.
Confidence: Official documentation, not tested · Sources: Claude Code securityWajo termsDots FAQCarly pricing
2026-10-08 · New agent
Added Carly from nine official sources. Current plans include Free limited tools, Personal $35/month with 200 credits, and Organization $35/seat/month. Phone/receptionist is a separately billed add-on. Privacy policy excludes Google, Microsoft and WhatsApp data from training and states deletion within 90 days after account deletion, with exceptions. Added a source-led Tab/Carly decision table and Tab card-link expiry.
Confidence: Official documentation and live pricing cards; not tested · Sources: PricingPrivacyMCP limits and permissions
2026-10-08 · Agent update
Rebuilt the Devin record from the pricing page and seven official docs. Cognition says customer data may train models by default, with a paid-plan opt-out it says enables zero data retention with model providers; retention lasts the customer relationship unless otherwise specified. Added CLI approval modes, admin security profiles, secrets scoping and the current Free/Pro/Max/Teams mechanics. No numeric quotas are published.
Confidence: Eight primary Cognition and Devin pages; documentation researched, not tested · Sources: PricingSecurity and privacy statementSecurity profilesCLI permissions
2026-10-08 · New agent
Added Tab with five official sources. Wallet features are limited beta; public service pricing and a fixed retention timetable are not established. Payment approval is documented, not tested. No blanket model-training exclusion is claimed.
Confidence: Five official pages; documentation researched, no account or hands-on test · Sources: Product overviewBeta terms and approvalsPrivacy and retentionSecurity statement
2026-10-08 · Agent update
Separated local approval modes from cloud command auto-run, added network and secret boundaries, and distinguished deleting transcripts from inactive snapshots (the documented window is now 30 days). Privacy-mode documentation differs; actual starting policy needs checking.
Confidence: Six primary Cursor pages; not tested or independently audited · Sources: Security overviewSecrets & NetworkLocal Run ModesRuntime choices
2026-10-08 · Agent update
Added default branch and workflow review controls, MCP/setup-step firewall exclusions, one-repository limits and Actions-minute costs. Free Copilot is not a free cloud-agent tier; training and retention terms remain unverified.
Confidence: Seven primary GitHub sources; no hands-on test or security audit · Sources: Risks and mitigationsFirewall scopeCloud agent overviewPlans
2026-10-08 · Agent update
Rechecked Google sign-in, GitHub repository scope, no-cost task allowances, personal-account paid eligibility, private-repository training scope and API approval controls. Local paid prices remain unverified.
Confidence: Six primary Google sources; documentation researched, not tested · Sources: Getting startedLimits and plansFAQJules API
2026-10-07 · Guide update
Added a Google-linked account teardown checklist, dated source links for all 52 agent records, and corrected Fo/Comma training and spending caveats. Older source checks remain labelled.
Confidence: Source-led checklist; no security audit or hands-on testing · Sources: Google linked-app helpWajo termsComma Data Use Policy
2026-10-07 · Comparison update
Expanded FAQ and policy evidence; corrected Wajo spending-control and credit assumptions, and checked Comma interactive Pro pricing.
Confidence: Primary vendor pages, no hands-on testing · Sources: Wajo termsComma pricingComma Data Use Policy
2026-10-07 · Comparison update
Rechecked six primary pages; added a decision table, Instinct vs Dots answer, use-case guidance and FAQs. Corrected the earlier gap on dots training settings.
Confidence: Primary vendor documentation; not tested · Sources: OpenAI dots FAQInstinct privacy policyMuse help
2026-10-06 · Comparison
Poke's published Free, $19 and $199 plans against Instinct's unpublished price, plus channels, action approvals and training opt-outs, from each vendor's own pages.
Confidence: Vendor pages, attributed; not tested · Sources: poke.com/pricinginstinct.com/terms
2026-10-06 · New agents
An AI phone assistant for iPhone, iPad and Mac that answers and makes calls from your own number. Its pricing page lists Pro, Premier and Prime plans by call hours, and a 7-day trial. TestingCatalog reports Mitra Link (assistants talking to each other, 5 October) and chat takeover (6 October); both come from announcements, not Mitra's help pages. Vendor description, not tested by us.
Confidence: Vendor site and pricing page checked; press used for newest features; not tested · Sources: getmitra.com/pricingtestingcatalog.com
2026-10-06 · Guide
Microsoft's $200 pack of 25,000 Copilot Credits, pay-as-you-go, credit rates per answer or action, who needs which licence, and the 125% enforcement limit.
Confidence: Vendor documentation, attributed · Sources: microsoft.comlearn.microsoft.com
2026-10-06 · New agents
OpenAI's coding agent across terminal, IDE, cloud and the ChatGPT desktop app. Its pricing page lists Free at $0, Go at $8, Plus at $20, Pro from $100 and Business at $20 per user a month billed annually, with usage shared with ChatGPT Work. No per-plan task allowance is stated.
Confidence: Vendor pricing page, docs and GitHub repository checked; not tested · Sources: developers.openai.comlearn.chatgpt.comgithub.com
2026-10-05 · New agents
A separate open-source iMessage assistant from Merit Systems, shown as an eve.dev template on 5 October. It uses a cloud browser and an encrypted vault, deploys to your own Vercel account (Pro plan required per its README) and asks for approval before purchases. It is a different project from Open Instinct by Maria Gorskikh. Security claims are the project's own.
Confidence: Project site, repository and eve.dev template page checked; not tested · Sources: openinstinct.shgithub.comeve.dev
2026-10-05 · New agents
An open-source, self-hosted personal agent you text on iMessage, published on GitHub under the MIT licence on 3 October 2026. Its README labels it version 0.1 and says it is not affiliated with Instinct. You supply your own model key, messaging line and hosting. Test counts and security claims are the author's own.
Confidence: Repository, README and docs checked; maintainer's X post used for context; not tested · Sources: github.comgithub.com/docs/KEYS
2026-10-05 · New agents
Open-source messaging infrastructure for developers who want one agent reachable over iMessage, WhatsApp, Telegram and SMS. Its pricing page lists a free tier, Pro at $25 a month and Business at $250 per line per month. The Spectrum core is MIT-licensed on GitHub. Uptime and deliverability claims are the vendor's own.
Confidence: Vendor pages, docs and public GitHub repository checked; not tested · Sources: photon.codesphoton.codes/docsgithub.com
2026-10-05 · New agents
A personal AI from Hark Labs with a browser-using agent called Handoff. It is on a waitlist, with no price published. The founder says launch is this week and the first 100,000 sign-ups get a paid plan free; the length and features of that plan are not stated. Handoff's benchmark claims are the vendor's own.
Confidence: Vendor pages and press coverage checked; launch timing is the founder's statement; not tested · Sources: hark.comtechcrunch.comruntimewire.com
2026-10-05 · New agents
An iPhone assistant that writes a daily briefing from mail, calendar and reminders and runs automations you describe in words. The vendor lists a 7-day free trial then $49 one-time, and says data stays on the phone while AI requests go to the provider you connect; it also lists Health, Music and Notes as coming soon.
Confidence: Vendor pages and App Store listing checked; vendor claims attributed; not tested · Sources: meethalo.appapps.apple.com
2026-10-05 · New agents
Perplexity's AI browser with an assistant. The vendor says the browser is free to download and lists Perplexity Pro at $20 a month and Max at $200 a month; which agent actions the free tier allows is not stated, so we do not claim it.
Confidence: Vendor pages checked; vendor claims attributed; not tested · Sources: www.perplexity.aiwww.perplexity.ai
2026-10-05 · New agents
An AI builder that makes websites and apps from a chat prompt. The vendor lists a free plan with a monthly token limit, a $25 a month plan and $30 per member for teams; token use grows with project size.
Confidence: Vendor pages checked; vendor claims attributed; not tested · Sources: bolt.newbolt.new
2026-10-04 · New agents
An AI email and calendar workspace for Gmail and Google Calendar. The vendor lists a free plan at $0 while it is new and a Teams tier as coming soon with no price; its agent drafts replies and waits for approval.
Confidence: Vendor pages checked; vendor claims attributed; not tested · Sources: hedwig-ai.com
2026-10-04 · New agents
Google's agentic development platform with a desktop app, CLI, IDE and SDK. The pricing page lists a $0 individual plan with basic weekly rate limits; the Google AI Pro price is not stated, so none is shown.
Confidence: Vendor pages checked; vendor claims attributed; not tested · Sources: antigravity.googleantigravity.google
2026-10-04 · New agents
GitHub Copilot cloud agent plans and edits code on a branch and opens pull requests; GitHub's plans page lists Pro at $10 a month and above for it. Lovable builds web apps from prompts and has a free daily credit grant; paid prices were not captured, so none is stated.
Confidence: Vendor pages checked; vendor claims attributed; not tested · Sources: docs.github.comgithub.comdocs.lovable.devlovable.dev
2026-10-04 · New agents
Rene (texting assistant, price not published), Town (work assistant with approval modes, free plan and $15 and $49 tiers), Lindy (Slack teammate from $29.99 per user a month) and Kiro (spec-driven coding agent, free to $200 a month by credits). Facts are from each vendor's own pages.
Confidence: Vendor pages checked; vendor claims attributed; not tested · Sources: rene.cowww.town.comwww.lindy.aikiro.dev
2026-10-03 · New agents · messaging
Ohai has a free version with Premium from $9.99 a month. Ollie bills by message: free 50 a month, $25 for 150, $100 for 1,000. Pally has a free plan and $25 a month. Orbits publishes no price we could find. TechCrunch's 3 October roundup was the lead; figures here come from each vendor's page, and the SOC 2 claim about Ollie is TechCrunch's, not confirmed on Ollie's pages.
Confidence: Vendor pages checked; the article is a lead only; vendor claims attributed; not tested · Sources: techcrunch.comwww.ohai.aiollie.aitryorbits.compally.com
2026-10-03 · New agents · messaging
Folk, Caddy, Fambot, Iris, Martin and Miso are added after checking each vendor's own pages. TechCrunch's article (3 October 2026) was used only as a lead. Prices shown are the vendors' own: Folk free with Pro at $9.99 a month, Iris from $11.99, Martin from $21 billed yearly; Caddy, Fambot and Miso publish no price we could find.
Confidence: Vendor pages checked; the article is a lead only; vendor claims attributed; not tested · Sources: techcrunch.comwww.folk.comirisai.appwww.trymartin.comcaddy.appfambot.commiso.com
2026-10-03 · New agents · Cursor, Jules
Cursor's cloud agents run in isolated VMs and open pull requests; its pricing page lists a free Hobby tier and $20 a month Individual. Google's Jules clones a GitHub repository to a cloud VM and returns a plan, diff and pull request; its page shows task limits but no dollar prices, so we list the price as unknown.
Confidence: Vendor pages only; vendor claims attributed; not tested · Sources: cursor.comcursor.comjules.google
2026-10-03 · New agent · Case
Case sells one persistent Linux desktop per agent that keeps logins, files and memory between sessions, driven over MCP from Claude Code, Codex or Cursor. Its site lists one plan at $29 a month with your own LLM key, and the code is on GitHub (licence listed as Other). We found no privacy policy or terms page. Developer infrastructure, not a finished assistant. Not tested.
Confidence: Vendor site and GitHub repository; no independent coverage found; vendor claims attributed · Sources: case.computergithub.com
2026-10-02 · Guide
What OpenAI's documentation says about the Agents API public beta: sessions, sandbox options, billing and data limits.
Confidence: Vendor documentation, attributed; public beta · Sources: developers.openai.comhelp.openai.comopenai.com
2026-10-02 · Guide
How Poke's documentation says recipes and custom MCP servers connect apps, and how to disconnect or refresh them.
Confidence: Vendor documentation, attributed · Sources: poke.compoke.com
2026-10-02 · Agent update · Lucas
Lucas's FAQ lists a free tier, Pro at $20/month, Max at $100/month and a seven-day reverse trial. Lucas now shows as having a free tier in the price filter. Numeric limits are not published.
Confidence: Vendor FAQ, attributed; not tested · Sources: meetlucas.ai
2026-10-02 · Agent update · Comma
Comma lists Value at $0/month, Pro at $80/month shown as billed yearly, and Enterprise as custom. Credits are priced at 1M per US$1 and do not roll over. Comma now shows as having a free tier in the price filter.
Confidence: Vendor pricing page, attributed; not tested · Sources: comma.surf
2026-10-02 · Agent update · Poke
Poke lists Free at $0, Pro at $19/month and Ultra at $199/month, billed monthly. No numeric limits are shown. Poke now shows as having a free tier in the price filter.
Confidence: Vendor pricing page, attributed; not tested · Sources: poke.com
2026-10-02 · New agent · Wabi
Wabi 2.0 is an iPhone messenger that handles plans and errands and builds small apps in chat, per its App Store listing. TechCrunch (29 Sep 2026) reports it is invite-only. Pricing is not published on the pages we read. Not tested.
Confidence: Vendor listing plus trade press · Sources: apps.apple.comtechcrunch.comwabi.ai
2026-10-02 · New agent · szn
szn is an iMessage-first personal agent from a former Siri and Apple Intelligence engineer. Its site describes calls, emails and follow-ups handled from one thread. Press from 27-28 Sep 2026 says it launches in October; price, countries and approval rules were not disclosed. Not tested.
Confidence: Vendor site plus trade press; launch date is press-reported · Sources: theszn.aiwww.ithinkdiff.comwww.macobserver.com
2026-10-02 · News · Anthropic
Claude Code adds mods: TypeScript functions that change how it behaves
Anthropic's blog says mods are small TypeScript functions, shipped inside plugins, that can rewrite a prompt, block or retry a tool call, approve or deny a permission request, or change the interface. They work in the CLI and desktop app. Anthropic warns mods are not sandboxed and run with the same machine access as Claude Code, so install only from trusted sources. On Team and Enterprise plans a built-in sec-default mod loads first to stop user-installed mods overriding permission deny rules. Vendor description, not tested by us.
Confidence: Primary source (Anthropic blog); the page shows no date in our fetch, trade press dates it 1-2 Oct 2026. Vendor claims attributed. · Sources: claude.com
2026-10-02 · New agent · Underdog · Funding news 2 Oct 2026
Underdog says its personal AI app (email, calendar, meeting notes, dictation, browser agent) runs its own models on your Mac and sends content to no cloud AI service. Its pricing page says it is free. a16z's 2 Oct 2026 post says it led first funding for Conway Research, Underdog's maker; round size was not disclosed. The site calls it an invite-only beta but also says to download it today on an Apple silicon Mac. Outside security audits are listed as under way. Not tested.
Confidence: Vendor pages plus a16z's own announcement (2 Oct 2026); funding size not disclosed; vendor claims attributed · Sources: underdog.aiunderdog.aiunderdog.aia16z.com
2026-10-01 · Site note
AgentComparison goes live with 16 documented agents
Sixteen agents at launch, one coding-agent guide and a comparison tool are live. Every record carries its sources and the date it was checked. Nothing has been hands-on tested.
Confidence: Not applicable
2026-10-01 · Guide
A documentation-led guide comparing what each vendor says about access, approval, retention and deletion.
Confidence: Vendor statements, attributed
2026-10-01 · Guide
Six jobs (coding, personal admin, errands, customer service, team workflows, building your own) mapped to the agents whose vendors document them.
Confidence: Vendor statements, attributed
2026-10-01 · Guide
Errand agent versus Task-based personal agent: approvals, spending, published pricing and what is still unknown.
Confidence: Vendor statements, attributed · Sources: wajo.aicomma.surf
2026-10-01 · Guide
Eight questions to ask before connecting accounts, with what 16 vendors say about approvals and data, and where they say nothing.
Confidence: Vendor statements, attributed
2026-10-01 · Guide
What agents are, how they differ from chatbots, how they work, what they are used for, what can go wrong, and where to start.
Confidence: Explainer; product statements attributed
2026-10-01 · Site note
Pricing filter added to the directory and comparison tool
Filter by free tier or open source documented, published USD price verified, usage-based, or price not verified. Unknown prices stay labelled as unknown.
Confidence: Not applicable
2026-10-01 · Guide
Which agents document a free tier or are open source, what is capped, and which prices we could not verify.
Confidence: Vendor statements, attributed
2026-10-01 · Guide
New app versus existing repository, how each lets you review changes, and published pricing, with a way to test them yourself.
Confidence: Vendor statements, attributed · Sources: docs.replit.comclaude.comdocs.devin.ai
2026-10-01 · Guide
How Microsoft and Salesforce price agent work (credits, conversations, licences) and what to ask about governance.
Confidence: Vendor pricing pages, attributed · Sources: www.microsoft.comwww.salesforce.com
2026-10-01 · Guide
Other agents in the directory grouped by what you want done, with published prices and caveats.
Confidence: Vendor documentation, attributed · Sources: help.manus.im
2026-10-01 · Guide
What LangChain's docs say LangGraph is, and when a framework or a ready-made agent is the shorter path.
Confidence: Vendor documentation, attributed · Sources: docs.langchain.com
2026-10-01 · Guide
What Intercom says counts as a billable Fin outcome and which handoffs are not charged.
Confidence: Vendor documentation, attributed · Sources: www.intercom.com
2026-10-01 · Guide
Who can run, edit and share, write-action approvals, connector constraints and API triggers, per OpenAI's help article.
Confidence: Vendor documentation, attributed · Sources: help.openai.com
2026-10-01 · Agent update · Claude Code
Anthropic's release notes for v2.1.287 (published 1 October 2026 18:00 UTC) say plugins can now modify deeper behavior (Claude Mods), add a built-in mod where a side agent flags things you or Claude might miss (for first-party sessions with telemetry on), and add URL prompts from MCP servers. It also lists a fix so a dangerous rm keeps its always-ask safeguard when the command also redirects output. If an MCP server stops connecting after the update, the notes say to add "bareElicitationCapability": true to its config entry.
Confidence: Primary source (vendor release notes) · Sources: github.com
2026-10-01 · Guide
What Meta's own safety post says about Sentinel approvals, credentials, purchases and training data, with the caveats it states.
Confidence: Vendor documentation, attributed · Sources: research.meta.ai
2026-10-01 · Guide
Devin's own billing documentation on Free, Pro, Max and Teams, full and flex seats, the $80 Teams minimum and on-demand credits.
Confidence: Vendor documentation, attributed · Sources: docs.devin.ai
2026-10-01 · Guide
How Anthropic's help article says usage is metered and the five habits it recommends.
Confidence: Vendor documentation, attributed · Sources: support.claude.com
2026-10-01 · Guide
How Replit's documentation says Agent work is charged: effort-based checkpoints, Plan Mode, plan features and spend controls.
Confidence: Vendor documentation, attributed · Sources: docs.replit.com
2026-10-01 · News · JetBrains
JetBrains opens Early Access for Air in its IDEs
JetBrains says the Air plugin (JetBrains Marketplace, or native in 2026.3 EAP builds) runs agents you already use, such as Codex, GitHub Copilot, Junie and Cursor, in parallel sessions inside the IDE. It says Air ships with no agents installed and sends no data until you sign in to an agent or a JetBrains AI plan. Free Junie Lite runs are offered during the EAP. Early access, vendor description, not tested by us.
Confidence: Primary source (vendor blog, 1 Oct 2026); early access, vendor claims attributed · Sources: blog.jetbrains.com
2026-10-01 · Guide
What Microsoft's documentation lists for admins: data policies, credit caps, audit logs, security scans, Agent 365 and the Lockbox exclusions.
Confidence: Vendor documentation, attributed · Sources: learn.microsoft.com
2026-10-01 · News · GitHub
GitHub Copilot adds computer use in public preview
GitHub's changelog says Copilot CLI and the Copilot app on macOS and Windows can now operate desktop apps: reading accessible content, clicking, typing, scrolling and dragging, including in software with no API or MCP integration. GitHub says Copilot asks for approval before controlling an app, you can review or reset apps you always allow, and organisation-managed settings can disable it. It is public preview. Vendor description, not tested by us.
Confidence: Primary source (GitHub changelog, 1 Oct 2026); public preview; vendor claims attributed · Sources: github.blog
2026-09-30 · Agent update · Claude Code
Claude Code 2.1.286: permission prompts show a count when requests stack up
Anthropic's release notes for v2.1.286 (published 30 September 2026 19:10 UTC) add an "N of M" count to stacked permission prompts, and fix several session-resume and login issues. Relevant if you review approvals often.
Confidence: Primary source (vendor release notes) · Sources: github.com
2026-09-30 · New agent · DoorDash
DoorDash says you can now text it what you want, and it finds options, builds the cart and checks out in the text thread, using your past orders. Access is a US waitlist beta. DoorDash also announced a connector for company AI tools to order team lunches. Announced 30 September 2026; the assistant is now in the directory.
Confidence: Vendor announcement plus trade press · Sources: about.doordash.comtechcrunch.comabout.doordash.com
2026-09-30 · New agent · Mindtrip
Mindtrip says its assistant learns traveler preferences, remembers them between trips and can run agents that watch prices and options. This comes from Mindtrip's own press release dated 30 September 2026, including its beta figures, which are unaudited. No price was stated. Now in the directory.
Confidence: Vendor press release only · Sources: www.prnewswire.com
2026-09-29 · Agent update · Claude Code
Claude Code 2.1.285: admins can limit API providers; web fetching can be turned off
Release notes for v2.1.285 (29 September 2026) add an allowedProviders managed setting to limit which API providers a machine may use, and an environment variable to disable the WebFetch tool.
Confidence: Primary source (vendor release notes) · Sources: github.com
2026-09-29 · New agent · Manus
Manus documents a rebuilt Manus 2.0 (Cascade engine, Manus Studio desktop workspace, Cloud Computer, Automations, Remote Control) and Cue, an invitation-only app where each agent has its own email, phone number, wallet and computer. Manus says its efficiency figures come from one tested configuration, not an average. InfoWorld reported the launch on 29 September 2026. Cue is now in the directory.
Confidence: Vendor documentation plus trade press · Sources: help.manus.imwww.infoworld.com
2026-09-29 · Agent update · Devin
Devin adds GPT-6.1 Sol in Devin Desktop and CLI
Cognition says OpenAI's GPT-6.1 Sol is live in Devin Desktop and Devin CLI. The cost and score figures in the post come from Cognition's own FrontierCode benchmark and are not independent.
Confidence: Primary source (vendor blog); benchmark claims are the vendor's · Sources: devin.ai
2026-09-29 · Agent update · Muse
Meta's announcement of 29 September 2026 says Muse, its personal agent for the US and Canada, now has skills and connectors for running a business, including Asana, Box, Canva, Dropbox, Figma, QuickBooks, Klaviyo, Notion, Shopify, Slack, Stripe, Zoom and Facebook and Instagram business accounts. Meta says nothing publishes, sends or spends without your approval, and that Muse is free for most of what people need, with paid subscription plans for more. Those are Meta's statements; we have not tested them, and it does not give plan prices.
Confidence: Primary source (vendor announcement) · Sources: about.fb.com
2026-09-28 · Agent update · Claude Code
Claude Code 2.1.284: Claude Sonnet 5.5 added as the default Sonnet model
The v2.1.284 release notes (28 September 2026) say Claude Sonnet 5.5 is now the default Sonnet model on the Anthropic API, with 1M context and $2/$10 per million tokens, cache reads $0.20 per million. These figures are Anthropic's, as stated in the notes.
Confidence: Primary source (vendor release notes) · Sources: github.com
2026-09-25 · Agent update · Replit Agent
Replit acquires Atta and adds interactive charts to Replit chat
Replit says it has acquired Atta and that you can ask Replit Agent to visualise data, with interactive charts in chat. Replit describes it as aimed at business analysis without writing SQL.
Confidence: Primary source (vendor blog) · Sources: replit.com