An incident.io teammate in Slack that answers who is on call, on incident.io's MCP server.

Yes, a remote one at mcp.incident.io/mcp. incident.io documents it as a way to query incidents, analyse alerts, check who is on call, manage escalations, and run operational analysis from any assistant that speaks the Model Context Protocol, and it names Claude Code, Cursor, Codex, VS Code, Claude Desktop, and ChatGPT among its clients. Chickpea connects to that same server over streamable HTTP and sends your API key as a bearer token. There is no incident.io endpoint written by Chickpea, and the connection cannot reach any other address.

incident.io describes itself as a software reliability platform. Teams use it to declare and run incidents, route alerts, hold on-call rotations and escalation paths, publish status pages, and write post-mortems afterwards. Its pricing page sells Basic, Team, Pro, and Enterprise plans, plus an on-call-only tier. A Chickpea teammate does not replace any of that. It reads and acts on what is already in your incident.io account, from a Slack channel, when someone asks.

incident.io is built around Slack and its own app does most of the work: declare an incident with /incident, run it in a dedicated channel, send alerts and updates into Slack, and run more than forty slash commands on the incident. Keep that app. A Chickpea teammate adds something else. It holds incident.io beside your other connected accounts in one channel, so a question about what broke, who is on call, and what the error tracker said can be one message rather than three tabs.

On incident.io's side, the MCP server sits on the Team, Pro, and Enterprise plans. incident.io's pricing page leaves it out of the free Basic plan, and the changelog entry that announced it calls it a public beta available to all paying customers. Past the plan, individual tools depend on products. incident.io marks the on-call tools as needing On-call and the document and telemetry tools as needing Investigations. A tool you are missing the product for still appears, and returns an error explaining why. Chickpea itself is Apache 2.0 and runs on your own Cloudflare account, so there are no seats and no metering on this side.

No, and incident.io has two of its own. @incident is an agent you tag in a Slack channel, a Teams tab, or the dashboard, and incident.io documents it as available to Pro and Enterprise customers. The Slack assistant is the same agent in Slack's assistant sidebar, and incident.io says it needs a Pro or Enterprise plan and a paid Slack plan. incident.io says both act with your own permissions and are attributed to you. A Chickpea teammate is different. It runs on your own deployment, it holds incident.io beside your other connectors, and on the API key path incident.io authenticates it as a service actor rather than as a person.

Yes, if the key allows it and your account has On-call. incident.io's tool table lists paging a user or an escalation path, acknowledging or declining a page, and putting someone on call for a period. Because a page wakes a person, the teammate says who it would reach and waits for a yes in the thread. Under that sits the gate: during a Slack turn Chickpea refuses any MCP tool whose name does not start with an unambiguous read verb unless your own message named that action.

incident.io's published tool table includes a few deletes, among them deleting an incident action, cancelling a maintenance window that has not started, and removing a schedule override. Each one is a write by name, so the read-verb gate applies: it is refused unless the message you just sent names it, and the teammate asks before anything that reaches other people. Past that, the key decides. A key without the permission for a thing gets an error from incident.io rather than doing it.

No. incident.io's status page tool drafts an update and nothing more, and incident.io explains why: publishing emails and texts everyone subscribed to the page, so a person always does that part. The teammate can hand you a draft in the thread. You publish it in incident.io.

Yes, and it is the point of incident.io. Its Slack app creates an incident with /incident, opens the incident channel, pulls in responders, sends updates over Slack, SMS, and email, and carries commands like /incident escalate and /incident whoisoncall. It also syncs a schedule into a Slack user group, so a mention like @platform-on-call always reaches whoever is on call now. A Chickpea teammate hears nothing from incident.io and posts nothing unprompted. It answers when someone asks, in the channel where the question came up. Both can be installed.

On your deployment, in a write-only Credential field outside the teammate record. It never enters the model's context, the tool arguments the model writes, or Slack. The model picks a connection by id and the key goes in at the last step on the way out. incident.io says API keys authenticate as a service actor rather than a specific user and do not expire until deleted, so create one for the teammate and delete it there when you disconnect. Disconnecting in Chickpea removes the stored secret and stops any schedule that depended on it.