AgentComparison
Clear method. Explicit evidence.

Find the agent that
fits the job.

What can it do, what access does it need, and what is still unknown? Compare AI agents without invented ratings or promises of full autonomy.

78Initial researched agents
5Groups by what the agent is for
2Products with limited pilots

Start with what the agent is for

Personal agents, different boundaries

Start with permissions and memory, not a leaderboard.

product · personal assistant

Instinct

A personal assistant you text or call for email, calendar and everyday tasks.

Instinct describes a personal assistant reached by text or call, connected to applications and devices.

Documentation researched

Pricing: Price not verified

Best fit: A messaging-first assistant connected to personal apps and devices.

product · personal assistant

Muse

Meta's personal agent for reminders, actions and file work.

Meta documents Muse as a personal agent for actions, reminders, ongoing conversations and file work.

Documentation researched

Pricing: Limited free use documented by Meta; paid price not established

Best fit: Personal reminders, ongoing tasks and artifact work.

product · personal assistant

OpenAI dots

OpenAI's personal assistant that uses plugins and shared ChatGPT memory.

OpenAI documents dots using permitted plugins and shared ChatGPT memory, with data-access and action controls.

Documentation researched

Pricing: Price not verified

Best fit: Personal connected-app work with shared ChatGPT context.

Choose the right comparison

One category is not one product.

A ready-made assistant, an enterprise agent builder and a developer framework solve different problems. We separate them before comparing capabilities.

Build a focused comparison →

How to use this directory

Want the buying questions first? Browse our guides to agent pricing, fit and permissions, grouped by the decision you need to make.

Pick the group that matches the job, open an agent, and read three things first: what it is documented to do, what access it needs, and what is still unknown. Each agent page links the vendor pages the facts come from, so you can check them yourself.

Is this a ranking?

No. We do not assign universal scores or rank agents. Vendor documentation supports the directory; the Bolt and Replit Agent records separately report limited pilot observations, with dates and limits.

Why do some agents show no price?

We only show prices the vendor publishes. When a vendor lists no price, or access is by waitlist or invite, the page says so instead of guessing.

How do we decide what to cover?

We add an agent only when a vendor page documents what it does, so the comparison has facts to work with. The approach follows Google's guidance on helpful, people-first content: say who wrote it, how it was made and why.

Coverage is a dated snapshot.

This directory does not claim to include every AI agent. Most records are documentation-only. Limited Bolt and Replit Agent pilots are not broad reliability tests or paid-plan cost comparisons. Pricing and access gaps are stated beside each agent.