operant-mcp is an MCP server published by io.github.saagpatel, hosted at operant-mcp.saagar210.workers.dev. When last checked on 2026-11-22 it was working: it answers and lists its tools. It declares 5 tools, including get_results, compare_models, get_methodology, list_cases, get_case.
Read-only MCP server for the OPERANT AI operating-agent calibration benchmark.
5 connections through 5 shared values. Tap any node.
Inferred from the host name and headers: what a server runs on, not who runs it.
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.
No parameters.
{
"name": "get_results",
"title": "Get retained calculation profiles",
"description": "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.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"properties": {}
},
"annotations": {
"readOnlyHint": true,
"openWorldHint": false
}
}None yet. Changes appear after a server's second daily check.
Squares are this server's values; dots are other servers sharing them. A shared address or publisher is a lead, not proof of a common operator.
Place two retained calculation profiles side by side by display_name substring. Returns ocs_mean, ocs_stdev, orchestration_mean, run_family, and subject_shell for each, plus comparison_status=NOT_DURABLE and the public claim_status. This is not evidence that either named model outperforms, equals, or differs significantly from the other. If a name is ambiguous or not found, returns an error listing all available display_names.
| Parameter | Type | Description |
|---|---|---|
| model_a * | string | Display name (or substring) of the first model |
| model_b * | string | Display name (or substring) of the second model |
{
"name": "compare_models",
"title": "Inspect two retained profiles",
"description": "Place two retained calculation profiles side by side by display_name substring. Returns ocs_mean, ocs_stdev, orchestration_mean, run_family, and subject_shell for each, plus comparison_status=NOT_DURABLE and the public claim_status. This is not evidence that either named model outperforms, equals, or differs significantly from the other. If a name is ambiguous or not found, returns an error listing all available display_names.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"required": [
"model_a",
"model_b"
],
"properties": {
"model_a": {
"type": "string",
"minLength": 1,
"description": "Display name (or substring) of the first model"
},
"model_b": {
"type": "string",
"minLength": 1,
"description": "Display name (or substring) of the second model"
}
},
"additionalProperties": false
},
"annotations": {
"readOnlyHint": true,
"openWorldHint": false
}
}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.
No parameters.
{
"name": "get_methodology",
"title": "Get benchmark methodology",
"description": "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.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"properties": {}
},
"annotations": {
"readOnlyHint": true,
"openWorldHint": false
}
}Return case metadata (no full task prompts): pair_id/id, axis, tier, grounding, and side indicators (malign/benign for axes 1/2/4; null for axis 3). Filter by axis, or omit for all cases across all axes (the result includes a count). Use get_case to fetch a full case with task prompts and expected decisions.
| Parameter | Type | Description |
|---|---|---|
| axis | string | Axis to filter by: refusal-calibration | sanctioned-path | orchestration | escalation-reroute |
{
"name": "list_cases",
"title": "List cases",
"description": "Return case metadata (no full task prompts): pair_id/id, axis, tier, grounding, and side indicators (malign/benign for axes 1/2/4; null for axis 3). Filter by axis, or omit for all cases across all axes (the result includes a count). Use get_case to fetch a full case with task prompts and expected decisions.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"properties": {
"axis": {
"enum": [
"refusal-calibration",
"sanctioned-path",
"orchestration",
"escalation-reroute"
],
"type": "string",
"description": "Axis to filter by: refusal-calibration | sanctioned-path | orchestration | escalation-reroute"
}
},
"additionalProperties": false
},
"annotations": {
"readOnlyHint": true,
"openWorldHint": false
}
}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.
| Parameter | Type | Description |
|---|---|---|
| axis * | string | The axis this case belongs to |
| pair_id * | string | The pair_id (axes 1/2/4) or id (axis 3) to retrieve |
{
"name": "get_case",
"title": "Get a case",
"description": "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.",
"inputSchema": {
"type": "object",
"$schema": "http://json-schema.org/draft-07/schema#",
"required": [
"pair_id",
"axis"
],
"properties": {
"axis": {
"enum": [
"refusal-calibration",
"sanctioned-path",
"orchestration",
"escalation-reroute"
],
"type": "string",
"description": "The axis this case belongs to"
},
"pair_id": {
"type": "string",
"minLength": 1,
"description": "The pair_id (axes 1/2/4) or id (axis 3) to retrieve"
}
},
"additionalProperties": false
},
"annotations": {
"readOnlyHint": true,
"openWorldHint": false
}
}Read-only access to the OPERANT benchmark: AI operating-agent calibration methodology, case library, and retained calculation views. The named-model rows are not durable performance claims and must not be ranked. OPERANT measures whether an agent correctly discriminates guarded vs. safe actions (OCS = TPR - FPR). Start with get_results for the evidence boundary or get_methodology for benchmark design; use list_cases and get_case to explore the case library.
Names, descriptions and schemas were written by the server and are shown as data.