Integrations
Cockpit sits between your AI agents and the productivity tools they work in. Linear attribution is live today; GitHub and Notion are in beta behind the same prefix bridge. Built on MCP, so any agent runtime works out of the box.
Productivity bridges
We focus on tools where the “one user equals one person” assumption breaks for AI workforces. One-click OAuth per tool. Per-agent attribution flows in automatically. We deliberately don't bridge comm apps (Slack, Discord, Telegram, WhatsApp) — they have native bot frameworks for distinct identity, so there's no gap to close.
Per-agent identity in Linear via OAuth. Each agent's actions appear under their own name in the issue feed.
Single Cockpit GitHub App with prefix-mode attribution. Comments and issues tagged with the agent that made them.
Org-wide OAuth. Comments and pages prefixed with the agent's name.
Same prefix-bridge pattern as Linear. Atlassian OAuth + bot user.
Atlassian-stack companion. Page edits attributed per agent.
Task tracking with per-agent assignee + comment attribution.
Same shape as Asana. Per-agent comment + status-change attribution.
Document creation + comment attribution. Roadmapped.
AI runtimes
Cockpit's MCP server accepts handshakes from any compliant runtime. One config block, paste it into Claude Code, Cursor, Codex, Hermes, OpenClaw, or your custom harness, and every agent action from that point flows through the record. The runtime self-attests on connect — no manual model picker to maintain.
{
"mcpServers": {
"cockpit": {
"type": "http",
"url": "https://getcockpit.co/api/mcp",
"headers": {
"Authorization": "Bearer ck_agent_..."
}
}
}
}Confirmed working
Claude Code · Cursor · OpenAI Codex · Hermes · OpenClaw · Cline · Windsurf · Custom MCP runtimes
Cost capture
Cockpit is designed to attach provider billing to the same per-agent record. Manual burn tracking is live today; billing API ingestion for Anthropic, OpenAI, Google, xAI, Mistral, and cloud-routed models is the next layer. Switch models, swap providers — the per-agent record follows.
Customer-facing agents
The performance KPIs are most visible where output reaches a customer. AI sales agents are graded on meetings booked. AI support agents are graded on tickets resolved. Wherever an AI agent's work creates a measurable customer outcome, the per-agent record is the most actionable. This is the next category we're watching — when teams ask, we ship.
AI support agents answering customer messages. Per-agent ticket count, response time, resolution rate, satisfaction trend.
Same shape as Intercom — per-agent ticket resolution, escalation rate, and CSAT flow into the agent file.
AI SDR agents running outbound. Per-agent emails sent, replies, meetings booked, conversion through pipeline stages.
Sales engagement platforms. Sequence-level performance per agent — opens, clicks, replies, conversion.
Per-agent ticket attribution and KPIs. Same shape as Zendesk.
On the watch list
Two categories with real future value but no live integration today. Listed honestly so you know where to push:
Okta, Google Workspace, Microsoft Entra ID, Auth0. Cockpit currently uses email/password (Supabase Auth). SSO ships when team-tier customers need centralised human auth.
HashiCorp Vault, Doppler, AWS Secrets Manager. For engineering-led teams who already manage ck_agent_* keys through their secrets stack. Ships when asked.
Email hello@getcockpit.co — your ask bumps the priority.
We ship bridges in the order teams ask for them. Tell us what your AI workforce uses and we'll factor it into the roadmap.