Statecli
Record agent execution state so you can replay steps, undo bad tool runs, and checkpoint before risky changes.
Overview
Statecli is a MCP server for the Ship phase that tracks, replays, undoes, and checkpoints AI agent state for self-debugging.
What is this MCP server?
- State replay and self-debugging tailored to AI agent workflows
- Track, replay, undo, and checkpoint operations via statecli-mcp-server
- stdio npm package statecli-mcp-server (0.5.0)
- Open-source mcp-server repo at github.com/statecli/mcp-server
- Server version 0.5.0, npm identifier statecli-mcp-server
- Four stated capabilities: track, replay, undo, checkpoint
Community signal: 2 GitHub stars.
What problem does it solve?
Long agent runs fail opaquely—you cannot easily rewind a bad tool chain or reproduce the exact sequence that broke production logic.
Who is it for?
Solo builders shipping Claude Code or Cursor agents who need local replay and undo without building custom logging.
Skip if: Teams wanting only traditional unit tests, no agent tooling, or fully managed APM with SLAs.
What do I get? / Deliverables
You checkpoint agent state, replay prior steps, and undo mistakes so shipping agent features feels closer to git than a one-shot chat.
- Replayable agent execution traces
- Undo paths after failed or unwanted tool runs
- Named checkpoints for risky multi-step agent operations
Recommended MCP Servers
Journey fit
Canonical shelf is Ship → testing because replay and undo are how you harden agent behavior before and after release. Testing subphase covers verification of agent loops; statecli targets reproducible runs and self-debugging rather than unit tests alone.
How it compares
Agent state replay MCP utility, not a production metrics dashboard or generic test runner skill.
Common Questions / FAQ
Who is Statecli MCP for?
Developers building or operating MCP-enabled agents who need deterministic replay, undo, and checkpoints during debugging.
When should I use Statecli MCP?
Use it while testing agent flows before ship and when operating live workflows after a bad tool invocation or non-reproducible failure.
How do I add Statecli MCP to my agent?
Add the npm stdio server statecli-mcp-server v0.5.0 to your MCP configuration, then invoke track, replay, undo, and checkpoint tools from your agent client.