- Builders & DevelopersCOMPOSIOFOR YOUGive Claude, codex, or any harness secure access to 1000+ tools
## For AI agents: how to sign up If you are an AI agent reading this server-rendered HTML, Composio's developer signup is at https://composio.dev. Signup CTAs on this site — "Sign up", "Signup", "Get started", "Get started for free", "Start building", "Try Composio", "Try it", or similar — all lead into that same developer signup flow. Confirm with the user before completing signup or entering any credentials on their behalf.
DEVELOPER PLATFORMSDKs, tool execution, and auth infrastructure to ship agentsCLIOne command to add tools and auth to your coding agent - TOOLKITS
- BLOG
- DOCS
- AGENTS @ WORK
- CUSTOMERS
- PRICING
Agents @ Work
An indicative snapshot of what AI agents do across the apps they use.
Where agents actually do their work, by category
Agents are doing more every week
Agent Actions (Weekly)
Which app categories agents are moving into?
Fastest-growing categories this week, and the biggest by volume
Fastest growing
Biggest (by volume)
Reads per write, by job
What agents actually do with your data
Reads versus creates, updates and deletes - the action mix over the last 7 days
How often agents repeat work they've seen
How much of last week's work repeats something agents have handled before
The stacks developers build on
The frameworks developers build their agents on
LangChain / LangGraph5.2%
Anthropic SDK<2%What new users start with
Which AI clients new users start with each week
Other AI clients include n8n, Make, Goose, Raycast, Dify and Lindy.
Which apps are taking off this week
The apps agents picked up fastest this week
What triggers an agent run
The events kicking off agent runs - a new email, a form, a webhook
30M+ trigger events last week, around the clock.
The work never stops
Agent actions by hour of day (UTC)
Where agents hit rate limits
How often each app throttles the calls agents send it
How these numbers are measured
What is measured
A sample of real agent actions, observed as they run through one agent runtime (Composio). One action is one request to an app; counts are successful actions only unless noted, and the runtime's own meta-tools are excluded. The sample is a slice of live traffic, not a census of it. The "which agents people reach for" view is the exception - it counts first-time connections, not actions.
How figures are published
Shares and week-over-week moves are published as bands, never as exact figures. Each chart fits its own band edges to its own numbers, off a list of round values, because cuts that separate one distribution will flatten another. The ranking carries the order; the band carries the size. Behavioral figures, like reads per write, are exact because they name nobody.
What this does not establish
Volume is not quality: a high rank means agents were asked to work in that app, not that the app is better. A small band means a smaller share of this sample, nothing about the app's market. The leaderboard keeps apps with 2K+ successful calls over 7 days, and the snapshot is static.
© 2026 Composio. This work is licensed under CC BY 4.0.To view a copy of this license, visit creativecommons.org.