Your coding agent
Reasons, writes and changes code
Explores the codebase, proposes fixes, runs commands and ships changes. You keep using the agents you already like.
For AI coding agents
The investigation layer your agents work through.
Claude Code, Codex, Cursor and other agents are great at reading and writing code. Mike Trace gives them somewhere durable to investigate: a shared case file that survives sessions, tools and handoffs, so switching agents doesn't mean starting over.
Positioning
Not a coding assistant. A place for evidence.
Mike Trace
Keeps the investigation
Preserves what was observed, tried, ruled out and verified, so each connected agent and every teammate builds on the same record.
Investigation continuity
Change the agent. Keep the investigation.
When work moves between connected coding agents, Mike Trace is designed to keep the investigation with the case, so the next investigator doesn't have to begin from zero. Only one agent actively drives an investigation at a time, and each new one starts from a clear handoff.
What stays with the case
- The current investigation state and latest handoff
- Evidence gathered so far
- Findings and the explanations still in play
- Contradictions that have been surfaced
- Prior attempts, including the ones that failed
Why it matters
Don't pay for the same investigation twice
Without a shared record, a new agent re-reads the code, re-runs the reproduction and re-tries approaches that already failed. With one, it starts from what is already known and spends its effort on what is still open.
Continuity applies to agents connected to Mike Trace over MCP, such as the clients listed below.
Supported clients
Bring the agents you already use.
Mike Trace connects over the Model Context Protocol (MCP), the open standard many agents already support.
- Claude Code Coding agentMCP · browser approval
- Codex Coding agentMCP · browser approval
- Cursor AI code editorMCP · browser approval
- Claude Desktop Desktop assistantMCP · OAuth sign-in
- Other MCP clients Any MCP-compatible agentMCP · OAuth or browser approval
Boundaries
Clear about what agents can and cannot do.
Agents connect with scoped, expiring access that you approve. They get what they need to investigate and nothing more.
Connected agents can
- Read the cases they have been given access to
- Record evidence, explanations, attempts and findings
- Take over an investigation and leave a handoff for the next agent
- Propose and document fixes and recovery steps
- Verify a fix and help close the investigation
Connected agents cannot
- See other accounts or projects they were not approved for
- Access your infrastructure, deployment or account settings
- Authorize risky recovery actions on your behalf
- Keep access forever: agent access expires and can be revoked