MCP install
labsly's MCP servers are free for AI agent developers. Use them in Claude Code, Cursor, Cline, Aider, or any MCP-compatible host.
None of these are published yet.
The servers are built and the tools below are real, but nothing has been pushed to PyPI —
pipx install will not find them today. Each one ships alongside
its app. Follow a product for its release date.
The servers
- systema-mcp — 27 tools for systems analysis: the wizard, CLD loops and critique, simulation, scenario comparison, assertions, and behaviour explanation. Built. Ships with Systema.
- labsly-atlas-mcp — Wardley canvas, AI critique, premortem, composer, and counterfactual verbs. Built. Ships with Atlas.
- Witness MCP — design critique tools for AI coding agents. Built. Ships with Witness.
- Dais MCP — audience modeling and persuasion analysis. Planned, not started. Follows Dais.
Install
Once a server is published, this is the shape of it. Systema shown as the example — the others follow the same pattern with their own package and command name.
Systema MCP
pipx install systema-mcp Add to Claude Code (~/.claude/config.json)
{
"mcpServers": {
"systema": {
"command": "systema-mcp",
"args": []
}
}
} Add to Cursor (.cursor/mcp.json)
{
"mcpServers": {
"systema": {
"command": "systema-mcp"
}
}
} Add to Cline / Aider
See per-host docs for MCP integration. The labsly servers follow the standard MCP protocol — any MCP-compatible host works.
Shared response envelope
All labsly MCP tools return a shared envelope so AI agents can reason about outputs consistently:
{
"result": "<tool-specific payload>",
"needs_revision": true | false,
"action_items": [
{ "severity": "high" | "medium" | "low",
"description": "<what to do>",
"rationale": "<why>" }
],
"pass_threshold": {
"score": 0.0-1.0,
"passed": true | false,
"explanation": "<what threshold means>"
},
"provenance": {
"tool": "<tool name>",
"model": "<model used>",
"confidence": 0.0-1.0,
"epistemic_source": "stated" | "inferred" | "synthesized" | "observed"
}
} Schema: labsly.com/schemas/persona-v1.json for the persona format used across the Suite.
Why free?
MCP servers are a distribution channel, not a revenue line. Every developer who installs a labsly MCP server is a potential Mac app customer. The Mac apps pay for the platform.