Reconstructing from raw material
- Scrolling back through logs and terminal output
- Re-reading long conversation histories
- Re-walking code paths a previous session already traced
- Re-trying approaches nobody recorded as failed
Model Context Protocol
Connect any MCP agent to Mike Trace.
MCP is the open standard that lets AI agents use external tools. Mike Trace speaks MCP, so your agents can investigate through it without custom plugins or glue code.
Context
Give agents investigation context, not just another log stream.
Raw logs and conversation history tell an agent what happened somewhere. A case tells it where the investigation stands. Structured evidence, case state and investigation history can reduce repeated discovery work compared with reconstructing context from scratch every session.
Starting from the case
- Structured evidence attached to the case
- The current state of the investigation and the latest handoff
- What has been attempted, disproved and verified
- Open questions and the explanations still in play
What agents get
Investigation tools, built for agents.
Once connected, an agent works with Mike Trace the same way a careful engineer would: read the case, gather evidence, record what was learned and hand off cleanly.
Read the case
Agents can open a case and understand where the investigation stands before touching any code.
Record as they go
Evidence, explanations, attempts and findings are written back to the case as the agent investigates.
Get oriented
Agents can ask Mike Trace what is known, what is still open and what a sensible next step looks like.
Hand off cleanly
When a session ends, the agent leaves a handoff so the next agent or teammate continues from where it stopped.
Verify and close
Agents help confirm the fix with fresh evidence and bring the investigation to a complete close.
Connecting
Standard sign-in. No pasted secrets.
Agents connect using standard OAuth sign-in or browser approval. You approve each agent for a specific project, and its access is scoped and expiring.
- 1
Add Mike Trace to your agent
Add Mike Trace as an MCP server in your agent or MCP-compatible client.
- 2
Sign in and approve
Sign in to your Mike Trace account in the browser and approve the agent for a specific project.
- 3
Start investigating
Your agent can now investigate cases in that project, with access that is scoped and expires.
Connection details and the exact endpoint for your account are shown in the Mike Trace Console.
Security
Designed for least privilege.
- Agents never receive your personal sign-in session
- Access is limited to the project you approve
- Access expires and can be disconnected or rotated at any time
- Every action an agent records stays attributed to that agent