Reelier MCP Server¶
MCP server for recording agent tool-call workflows and replaying them deterministically at zero tokens. Agents make claims - Reelier writes receipts. Record an agent's workflow once, replay it exactly, and diff runs to catch drift.
What It Does¶
- Workflow Recording - Capture every tool call, parameter, and response in an agent session
- Deterministic Replay - Replay recorded workflows exactly - no LLM, no tokens, no variation
- Drift Detection - Diff current runs against recorded baselines to detect changes
- Zero Token Cost - Replay costs nothing because it's purely deterministic
Business Operator Use Cases¶
| Use Case | Value |
|---|---|
| QA Automation | Record a successful agent workflow once, replay for every deployment |
| Agent Auditing | "What exactly did the agent do during that incident?" → full replay |
| Cost Optimization | Replay deterministic workflows instead of burning tokens on repeated tasks |
| Compliance | Immutable record of every agent action for SOC2/ISO audits |
Installation¶
git clone https://github.com/maximehoule/reelier
cd reelier
npm install
Configuration¶
Add to your MCP client config:
{
"mcpServers": {
"reelier": {
"command": "node",
"args": ["path/to/reelier/index.js"],
"env": {
"REELIER_STORAGE_PATH": "/path/to/workflow-recordings"
}
}
}
}
Tools Provided¶
start_recording- Begin capturing agent workflowstop_recording- End capture and save workflowreplay_workflow- Deterministically replay a saved workflowdiff_workflows- Compare two workflow runs, identify driftlist_recordings- Browse saved workflow recordings
Integration with CorpusIQ¶
CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.
This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.
Limitations¶
- 0 stars, brand new - Created July 22, 2026. Experimental.
- Deterministic only - Replay is exact. Can't adapt to changed environments or APIs.
- Storage growth - Long workflows with large responses consume disk space.
- No scheduling - Manual trigger only. No cron/CI integration yet.
Operator Verdict¶
★★★ Critical infrastructure for agent operations at scale. This is one of the first MCP servers addressing the "how do we know what our agents are doing" problem. Deterministic replay for compliance and cost savings is genuinely innovative. Watch for CI/CD integration and scheduling features. This could become standard equipment for any team running >3 agents in production.