AgentComparison
Microsoft · platform · agent-building platform

Microsoft Copilot Studio

Microsoft's platform for building and governing agents on business data.

Group: Build-your-own platforms and infrastructure

Microsoft describes Copilot Studio as a no-code platform for creating and managing custom agents, workflows and apps on business data, published across channels.

Documentation researched, not testedChecked 30 October 2026 (billing rates and replenishment re-read 3 November 2026)Vendor documentation, not a test

The decision in plain English

Compare governance, connectors and operating cost with other builders, not consumer assistants.

Best fit

Teams building business-data agents for several channels.

Poor fit

Comparing it as a ready-made personal assistant.

Documented capabilities

Vendor statements, not verified task outcomes.

  • Business-data connections Source · checked 2026-10-22
  • Natural-language agent building Source · checked 2026-10-22
  • Multi-channel publishing Source · checked 2026-10-12

Permissions and human control

  • User authentication and channel support must be selected; No authentication permits anyone with the link.
  • Event triggers use maker credentials; autonomous actions require working maker authentication, which can expose data/actions to published users.
  • Makers can require confirmation for tools that modify data. Admin data policies can block connectors, knowledge, HTTP, channels and event triggers.

Documented controls do not prove that an agent always follows them.

Limits and open questions

  • Trigger payloads can consume metered usage; unattended does not mean free.
  • Maker-authenticated triggers can expose maker-account data/actions.
  • Publishing and auth changes must be tested in the intended environment and channel.
  • Product-specific retention/training controls were not established in this feature check.

Privacy and security

Detailed privacy, retention and security controls were not established by the sources in this overview. Verify the current vendor policy before connecting sensitive data. No certification or audit is claimed.

Feature-by-feature: what is documented and what is not

Official sources checked 30 October 2026 (billing rates and replenishment re-read 3 November 2026). A documented feature is a vendor statement, not proof of success.

FeatureEvidence statusDetail and limit
Building blocksDocumentedGenerative orchestration selects configured topics, tools, agents and knowledge. It does not establish successful task completion. Source
Event triggersRequires generative orchestrationOutside events and recurrence can start work without a user prompt. Trigger availability depends on environment settings and data policies. Source
Trigger credentialsMaker account riskTriggers use maker credentials. Autonomous authenticated actions also need working maker authentication; published users may invoke data/actions using that access. Source
Human confirmationConfigurableMakers can require confirmation for tools that modify data. Do not assume all changes prompt merely because the agent is published. Source
Data policiesAdmin configuredPolicies can block knowledge sources, connectors, HTTP, skills, channels and triggers. Data cannot be shared across connectors in different policy groups. Source
User authenticationChannel dependentNo authentication lets anyone with the link interact. Microsoft authentication supports Teams/Microsoft 365; other authenticated channels need manual setup. Changes take effect after publication. Source
Usage and external channelsLicensed and meteredThe pricing page distinguishes included internal Microsoft 365 Copilot rights from standalone external channels. Packs, pay-as-you-go and prepurchase exist; Azure and overage coverage need review. Source
Background billingNot free because unattendedTrigger payloads count toward consumption when they trigger actions or responses, including recurring payloads. Source

How to start with Microsoft Copilot Studio narrowly

  1. Use a test environment and one read-only knowledge task.
  2. Check channel authentication, connector data-policy groups and whose credentials each action uses.
  3. Preview approved outputs before publication; require confirmation for risky tools.
  4. Add event triggers only after reviewing maker-account exposure, scope and usage budget.

Official product/setup source. These are documentation-based steps, not a hands-on setup test.

Common questions about Microsoft Copilot Studio

What is Microsoft Copilot Studio for?

Teams building business-data agents for several channels. The vendor describes it as: Microsoft describes Copilot Studio as a no-code platform for creating and managing custom agents, workflows and apps on business data, published across channels.

Who should look elsewhere?

Comparing it as a ready-made personal assistant.

How much does Microsoft Copilot Studio cost?

Microsoft lists a tenant-wide license at USD $200/pack/month for 25,000 Copilot Credits; also a pay-as-you-go meter (requires an active Azure subscription linked to the environment) and a pre-purchase plan of Copilot Credit Commit Units with up to 20% savings and automatic pay-as-you-go backstop when prepaid credits run out. The pay-as-you-go meter is priced at $0.01 per Copilot Credit, billed in arrears at month end, and requires no up-front licence commitment; where credit packs are also purchased, consumption counts against the meter only once those packs are exhausted. Packs are whole units, billed annually, and unused credits do not roll over (Microsoft Copilot Studio Licensing Guide, June 2026, read 2 November 2026). Actions/responses consume varying credits by usage. Microsoft 365 Copilot ($30/user) includes different, separate rights. Prepaid capacity-pack credits are replenished at the start of every monthly billing period (the subscription's own billing cycle, not a fixed calendar date); standalone prepaid users who exhaust credits lose Copilot Chat access until replenishment unless pay-as-you-go is active as a safety net (Microsoft Learn capacity-packs doc, read 3 November 2026).

What can it access and what needs my approval?

User authentication and channel support must be selected; No authentication permits anyone with the link. Event triggers use maker credentials; autonomous actions require working maker authentication, which can expose data/actions to published users. Makers can require confirmation for tools that modify data. Admin data policies can block connectors, knowledge, HTTP, channels and event triggers.

What happens to my data?

We did not establish a vendor statement on retention or training in the sources we read. Ask the vendor before connecting sensitive data.

How much does a knowledge-grounded Copilot Studio answer cost?

Microsoft's Standard-harness base feature rates (read 3 November and re-read 7 November 2026): classic answer 1 Copilot Credit, generative answer 2, agent action 5, tenant graph grounding for messages 10, agent flow actions 13 per 100 actions, content processing 8 per page. Features stack within one prompt - Microsoft's own example: tenant graph grounding plus a generative answer totals 12 credits for one complex response before any reasoning-model token charge or other metered activity. At the $0.01 pay-as-you-go rate that is about $0.12 for those base features, not a universal all-in answer cost; a $200 pack of 25,000 credits covers about 2,083 such base-feature combinations arithmetically, before extra consumption. Employees licensed for Microsoft 365 Copilot are not charged for included employee-facing usage, subject to conditions and fair-use limits.

Source

Do reasoning models and testing add Copilot Studio charges?

Microsoft Learn, read 7 November 2026, distinguishes billing paths. For the Standard harness, reasoning-model use adds Text and generative AI tools (premium) at 10 Copilot Credits per 1,000 billed tokens on top of the operation's feature rate. Input and output tokens, including reasoning, count; cache writes count as input, but cache reads do not. Multiple model calls for one message each count. Microsoft's illustrative 14,100-billed-token response therefore adds 141 credits before feature charges, not a measured task cost from our test. Long instructions, retrieved knowledge and conversation history can all affect that total. Bring-your-own models are billed separately. Do not carry the Standard-harness 2-credit generative-answer figure over to the GitHub Copilot harness: its overview says LLM tokens, tools and harness usage consume credits, including building, previewing, testing and evaluations. Testing is not universally free: AI Builder's rules separately say prompts/models in agent flows consume credits even from a test panel or flow designer. Check the actual harness, model, trigger and usage report before estimating a run. We have not tested either billing path.

Source · Source · Source

Can I see and limit Copilot Studio credit spending?

Microsoft's 4 November 2026 update describes credit visibility during building/testing and workflow reports by run and node. Copilot Credit cost management in the Microsoft 365 admin center is labelled preview. It describes user-based spending policies at tenant/group level and agent-based limits at environment level, plus alerts. Verify which controls your tenant can actually use before treating them as a spending boundary; a dashboard or alert alone is not a universal hard cap. This does not mean every Copilot Studio customer has automatically enabled pay-as-you-go, and it is separate from future Copilot Business CSP billing changes. We have not tested these controls.

Source

Has AgentComparison tested it?

No. This record is built from the vendor's own documentation, checked 2026-10-30. Vendor claims are attributed, not confirmed.

How to evaluate Microsoft Copilot Studio yourself

  1. Pick one low-risk task that matches its documented capabilities: business-data connections.
  2. Write the expected output and what counts as failure before you start.
  3. Connect the minimum access needed, and check approval settings first.
  4. Run it twice and compare, then review every action it took.
  5. Check cost after any free allowance. Unknown price is not free.

Related

Documented use cases: Repeatable team and business workflows; Building your own agent.

Sources and history

Change history

Documentation overview expanded with current pricing/access sources. No hands-on results claimed.

Report a correction