Documentation

Integrations

Supported coding agents and adjacent systems AgentPM can capture, observe, or connect to evidence.

Supported and planned integrations

AgentPM is strongest when it can find the local transcript and normalize it into the shared conversation model.

Claude Code

Supported. AgentPM captures Claude Code project JSONL sessions, subagent transcripts, tool use, model usage, cwd metadata, skills, commands, and project memories where available.

Open

Codex

Supported. AgentPM captures Codex session JSONL files, tool calls, turn ids, cwd records, model settings, token usage records, and history-derived conversations.

Open

Cursor

Supported. AgentPM captures Cursor agent transcript JSONL files, user queries, tool use, selected metadata, and project grouping from Cursor project paths.

Open

Grok

Supported. AgentPM captures Grok session and log JSONL files where present, then normalizes them into the same conversation model as other coding agents.

Open

OpenCode

Supported. AgentPM captures OpenCode sessions from the local SQLite store, materializes them into collector spool evidence, and normalizes them into the same conversation model as other coding agents.

Open

Gemini CLI

Planned dedicated support. AgentPM recognizes Gemini as an agent family, but dedicated Gemini CLI collection is still a future integration. Use generic JSONL capture only when your setup produces compatible logs.

Open

Git

Evidence layer. AgentPM connects conversations to branch, commit, and file-change evidence so teams can understand how local agent work moved toward a shipped diff.

Open

GitHub

Evidence layer. AgentPM can connect conversation evidence to pull request and branch trajectories, helping reviewers understand what happened before the final PR summary.

Open

MCP

Observed in sessions. AgentPM surfaces MCP server and tool usage where captured by the underlying agent transcript, especially in project views and conversation evidence.

Open

Future integrations

Extensible. New integrations should define where transcripts live, how collection discovers them, how conversations are normalized, and what metadata becomes searchable evidence.

Open