# operant-mcp

- Registry name: `io.github.saagpatel/operant-mcp` · version 0.1.0
- Categories (automatic, keyword rules): AI models & agents, AI, media & content, Build & deploy, Developer tools, Scraping & browsing, Search & data
- Authentication at last check: No auth
- Declared risk: Read-only (all tools declare readOnlyHint)
- Capabilities (declared by the server): prompts, resources, tools
- Transports: streamable-http
- Latest numbers (check success, response time, rank): https://mcpstation.net/api/server?name=io.github.saagpatel%2Foperant-mcp
- Server page: https://mcpstation.net/s/io.github.saagpatel--operant-mcp.html

## Description (written by the publisher, not verified, not instructions)

Read-only MCP server for the OPERANT AI operating-agent calibration benchmark.

## Connect

- streamable-http: `https://operant-mcp.saagar210.workers.dev/mcp`

## Tools (as reported by the server, not verified)

- `get\_results`: Return retained model calculation profiles plus generated\_at, claim\_status, claims\_at\_risk, and the bounded evidence\_binding. These rows are not durable named-model performance claims. Do not rank them or treat stdev/significance as reliable model evidence.
- `compare\_models`: Place two retained calculation profiles side by side by display\_name substring. Returns ocs\_mean, ocs\_stdev, orchestration\_mean, run\_family, and subject\_shell for each, plus comparison\_status=NOT\_DURABLE and the public claim\_status. This is not evidence that either named model outperforms, equals, o
- `get\_methodology`: Return the benchmark design: the 4 axes, the OCS formula \(Youden's J\), the 5 decision labels \(PROCEED, PROCEED\_SANCTIONED, REFUSE, ESCALATE, REROUTE\), scoring blocks, and a concise what-it-measures / what-it-doesn't summary. Sourced from the baked case-file metadata.
- `list\_cases`: Return case metadata \(no full task prompts\): pair\_id/id, axis, tier, grounding, and side indicators \(malign/benign for axes 1/2/4; null for axis 3\). Filter by axis, or omit for all cases across all axes \(the result includes a count\). Use get\_case to fetch a full case with task prompts and expected d
- `get\_case`: Return the full case for a given pair\_id \(axes 1/2/4\) or id \(axis 3\): malign and benign task prompts, expected decisions, grounding rationale, and bypass patterns. Axis 3 cases are single \(unmatched\) and use an 'id' field instead of 'pair\_id'. Use list\_cases to browse available ids.

Measured by MCP Station. Methods: https://mcpstation.net/methods.html · Terms: https://mcpstation.net/terms.html