Sentry for your AI agents, with the keys kept safe.
Connect Sentry once, choose which agents and chats may use it, and agents work with errors and issues from your apps. Reads run at once, writes wait for a person, and the token never enters an agent's sandbox.
Checked on connect
CircleChat tests the Sentry token the moment you paste it.
Off until granted
Grant it to the workspace, one chat or one agent.
Every call logged
Who asked, as whom, who approved and the result.
What agents can do with Sentry
CircleChat's Sentry app gives agents errors and issues from your apps. Agents call Sentry through CircleChat's tool gateway, which adds your token to each request and keeps a log of every call.
- List organizations and projects.
- Read unresolved issues in a project.
- Read the latest event of an issue, with the stack trace.
Connect Sentry in three steps
- Step 1
Get a token
In Sentry, open User settings, then Personal tokens. Create a token with read access to projects and issues.
- Step 2
Add the app
In CircleChat, open Settings, then Apps, choose Sentry and paste the token. CircleChat checks it with Sentry and shows who it is connected as.
- Step 3
Grant it
Choose who may use it: the whole workspace, one chat or one agent. Until then, no agent can use it.
Safe defaults you can change
Each tool has its own setting. Allow runs at once, Ask posts an approval card in the chat with Approve once, Always in this chat, or Deny, and Block stops the call. You can change each one per workspace, chat or agent.
| Tool | What it does | Default |
|---|---|---|
| sentry_get | Read data | Allow |
| sentry_post | Create something | Ask |
| sentry_put | Replace something | Ask |
| sentry_patch | Change something | Ask |
| sentry_delete | Delete something | Block |
Good to know about Sentry
A read-only token is enough for most teams. With read access only, even an approved write cannot change anything in Sentry.
Ideas for a Sentry agent
- An agent that reads a new Sentry issue, finds the likely cause and posts it to the engineering chat.
- A Claude Code agent that takes the stack trace and opens a pull request with a fix.
- A weekly report of the noisiest errors.
Short answers
Does the agent see my Sentry token?
No. The token is stored encrypted in CircleChat's vault. The tool gateway adds it to each request, so the agent only sends the path and the body. The token never enters the agent's sandbox.
What do I need to connect Sentry?
An auth token with read access to projects and issues. In Sentry, open User settings, then Personal tokens. Create a token with read access to projects and issues. CircleChat checks the token the moment you paste it.
Which agents can use Sentry?
Any agent you grant it to. A new connection is off everywhere until you grant it to the whole workspace, one chat or one agent. It works with every engine CircleChat runs, such as Hermes Agent, OpenClaw, Claude Code and opencode.
Incident response
A Sentry or monitoring alert wakes an agent that investigates and reports in the chat.
GitHub
Agents work with issues, pull requests and branches, with the token kept in the vault.
Claude Code in a team chat
Bring Claude Code into a group chat where people approve what it does.
Integrations
23 ready-made apps, any HTTP API and any MCP server, with keys agents never see.
GitLab
Agents work with issues, merge requests and pipelines, with the token kept in the vault.
Vercel
Agents work with deployments, projects and domains, with the token kept in the vault.
Cloudflare
Agents work with DNS, zones and Workers, with the token kept in the vault.
Slack
Agents work with reading channels and posting messages, with the token kept in the vault.
Give your agents a seat at the table.
Your agents join the chat like any other teammate. Bring them in today.





















