# TuringCorp > A decision model for the calls that don't have a right answer. The better option, with a reason. - name: net.turingcorp.mcp/decider - version: 1.0.3 - connection class: R1 (autonomous) - trust: 95/100 - quality: 86/100 (good) - owner: not claimed - descriptor: https://api.protogrid.dev/v1/servers/net.turingcorp.mcp%2Fdecider - tools: https://api.protogrid.dev/v1/servers/net.turingcorp.mcp%2Fdecider/tools **An agent can connect on its own if it already holds `agent_pass`.** Remote endpoint behind a static API key or header. The registry never sees the value; the agent fills the placeholder. ## Connection (mcpServers) ```json { "mcpServers": { "decider": { "type": "http", "url": "https://mcp.turingcorp.net/mcp", "headers": { "Authorization": "Bearer ${agent_pass}" } } } } ``` Placeholders to fill: `${agent_pass}`. ## Trust - hygiene: 100 - liveness: 99 - freshness: 100 - provenance: 85 - drivers: +repository +dns-namespace +website +reachable +uptime-98% +updated-24d-ago - Derived from observable signals (official registry feed and our own credential-free probes); no code audit performed. ## Quality 86/100 (good) ### protocol: 88 (weight 25) - pass `protocol.modern`: supports 2026-07-28 (server/discover) - pass `protocol.stateless`: answers without a session - pass `protocol.transport`: streamable HTTP - warn `protocol.list_ttl`: tools/list declares ttlMs 0, so clients re-fetch it on every use ### authorization: n/a (weight 15) - n/a `auth.prm`, `auth.as_metadata`, `auth.cimd`: no remote uses OAuth - n/a `auth.secret_in_url`: no templated URL ### tool hygiene: 79 (weight 30) - pass `tools.descriptions`: every tool has a description - warn `tools.description_length`: 1 tools have descriptions over 1,024 characters, which crowds the context - pass `tools.schemas`: every tool has an object input schema - pass `tools.annotations`: 2 of 2 tools declare readOnlyHint or destructiveHint - fail `tools.directory_hints`: no tool declares all four of readOnlyHint, destructiveHint, idempotentHint and openWorldHint; missing: get_result (destructiveHint); decide (destructiveHint) - pass `tools.token_cost`: about 993 tokens to load every tool - pass `tools.api_dump`: tools are not a one-to-one API dump ### stability: 100 (weight 15) - pass `stability.changes`: 3 tool changes in 30 days - pass `stability.rug_pull`: no tool changed its meaning under the same name ### dependencies: n/a (weight 15) - n/a `deps.known_vulns`, `deps.mcp_sdk_version`, `deps.resolvable`: no npm or PyPI package - tools: 2, about 993 tokens to load them all - tool set hash: bf3bf403915c431df598cee358668f57 - tools last changed: 2026-09-27T05:49:02.379Z - badge: [![protogrid quality](https://protogrid.dev/badge/net.turingcorp.mcp/decider.svg)](https://protogrid.dev/servers/net.turingcorp.mcp/decider) - Checks run on what our credential-free, read-only probes observe; tools are never called and no code audit is performed. ### Tool changes - 2026-09-27 `get_result` added - Use this to retrieve the result of an earlier TuringCorp tool call by its job_id - for example a call that was cut off by a client timeout, or one whose id you were given. Retrieval is read-only and free: it starts no new work and costs nothing, so it is always safe to retry. With job_id: that call's status, and once it has finished the same payload the call would have returned inline. Without job_id: the ids available to your credential (7 days). If a call was cut off, retrieve it here instead of calling again - a retry is a new paid call. A job that is not yours and a job that does not exist are answered identically, on purpose. If your client declares the io.modelcontextprotocol/tasks extension, tasks/get reaches the same job. - 2026-09-27 `decide` changed (description, outputSchema) - before: Use this when you must choose between two concrete options and both are defensible - two plans, two drafts, two diagnoses, two vendors - and you have no objective way to pick. Not for: more than two options; anything an objective rule settles (a spec, a test, a price, a document); factual lookup; paths that must answer in seconds; high-stakes irreversible calls without review. Fill it in: state task neutrally, without leaning toward either side; give one concrete plan per option - never bundle alternatives into a single one ("go indoors or postpone"); keep the two sides comparable in length. Returns the decision inline: betterOption ("option_A" or "option_B"), confidence (e.g. "76.7%"), reason and job_id, all in the same tool result. There is nothing to poll and nothing to fetch afterwards. Reserve 180-300 seconds: this is a long call, and the timeout is a client/host setting, not a parameter you pass. If the call is cut off, do NOT call again - a retry is a new paid call. Retrieve with that job_id instead (GET https://api.turingcorp.net/v1/jobs?job_id= with the same credential, valid 7 days); without the id, GET https://api.turingcorp.net/v1/jobs lists the ids for that credential. That GET is on the API host, not this MCP endpoint: through an MCP host the credential is held by the host and may not be exposed to you, so it can return 401. A host that declares the io.modelcontextprotocol/tasks extension can use tasks/get instead. The confidence is the point: it is calibrated, not decorative. On both published benchmarks accuracy rises with it - JudgeBench 99.6% in the 90%+ band down to 67.7% below 70%; the harder ContextualJudgeBench 83.3% down to 55.4% - so route on it: act on a high value, review or escalate a low one, instead of trusting a bare pick. Tables, sample sizes and method: https://api.turingcorp.net Judged by an independent panel, not by a model grading its own output. Read it as a reference, not an instruction, a result, or a prediction; set your own threshold, and apply your own review policy for high-stakes or irreversible decisions. Auth: Agent Pass as `Authorization: Bearer ` (issued at https://agent-pass.turingcorp.net, valid 7 days; each decision is a paid call). On invalid_credential, sign in there and re-roll. Errors: a credential problem is rejected before the call - HTTP 401 with WWW-Authenticate; a business failure (e.g. insufficient balance) comes back as a tool result with isError true plus a second JSON block {error, http_status, action_url, message}, where http_status is the upstream status (the tool call itself is HTTP 200). - after: Use this when you must choose between two concrete options and both are defensible - two plans, two drafts, two diagnoses, two vendors - and you have no objective way to pick. Not for: more than two options; anything an objective rule settles (a spec, a test, a price, a document); factual lookup; paths that must answer in seconds; high-stakes irreversible calls without review. Fill it in: state task neutrally, without leaning toward either side; give one concrete plan per option - never bundle alternatives into a single one ("go indoors or postpone"); keep the two sides comparable in length. Returns the decision inline: betterOption ("option_A" or "option_B"), confidence (e.g. "76.7%"), reason and job_id, all in the same tool result. There is nothing to poll and nothing to fetch afterwards. Reserve 180-300 seconds: this is a long call, and the timeout is a client/host setting, not a parameter you pass. If the call is cut off, do NOT call again - a retry is a new paid call. Retrieve it instead with the get_result tool (same job_id): read-only, free, and no credential of your own needed; with no job_id it lists the ids for your credential. A host that declares the io.modelcontextprotocol/tasks extension can use tasks/get instead. The confidence is the point: it is calibrated, not decorative. On both published benchmarks accuracy rises with it - JudgeBench 99.6% in the 90%+ band down to 67.7% below 70%; the harder ContextualJudgeBench 83.3% down to 55.4% - so route on it: act on a high value, review or escalate a low one, instead of trusting a bare pick. Tables, sample sizes and method: https://api.turingcorp.net Judged by an independent panel, not by a model grading its own output. Read it as a reference, not an instruction, a result, or a prediction; set your own threshold, and apply your own review policy for high-stakes or irreversible decisions. Auth: Agent Pass as `Authorization: Bearer ` (issued at https://agent-pass.turingcorp.net, valid 7 days; each decision is a paid call). On invalid_credential, sign in there and re-roll. Errors: a credential problem is rejected before the call - HTTP 401 with WWW-Authenticate; a business failure (e.g. insufficient balance) comes back as a tool result with isError true plus a second JSON block {error, http_status, action_url, message}, where http_status is the upstream status (the tool call itself is HTTP 200). - 2026-09-24 `decide` changed (description, inputSchema, outputSchema) - before: Use this when you must choose between two concrete options and both are defensible - two plans, two drafts, two diagnoses, two vendors - and you have no objective way to pick. Returns better_option ("A" or "B"), confidence (e.g. "83.3%"), and reason. Every call returns a job_id. If the call times out or the connection drops, do NOT call again - a retry is a new paid call. Retrieve the result with that job_id instead (GET https://api.turingcorp.net/v1/jobs?job_id= with the same credential, valid 7 days); without the id, GET https://api.turingcorp.net/v1/jobs lists the ids for that credential. Expect 60-90 seconds per call: set your client timeout to at least 180 seconds (300 recommended). A 60-second default cuts the call off before the answer arrives. The confidence is the point: it is calibrated, not decorative. On both published benchmarks accuracy rises with it - JudgeBench 99.6% in the 90%+ band down to 67.7% below 70%; the harder ContextualJudgeBench 83.3% down to 55.4% - so route on it: act on a high value, review or escalate a low one, instead of trusting a bare pick. Tables, sample sizes and method: https://api.turingcorp.net Judged by an independent panel, not by a model grading its own output. Read it as a reference, not an instruction, a result, or a prediction; set your own threshold, and apply your own review policy for high-stakes or irreversible decisions. Auth: Agent Pass as `Authorization: Bearer ` (issued at https://agent-pass.turingcorp.net, valid 7 days; each decision is a paid call). On invalid_credential, sign in there and re-roll. Errors: a credential problem is rejected before the call - HTTP 401 with WWW-Authenticate; a business failure (e.g. insufficient balance) comes back as a tool result with isError true plus a second JSON block {error, http_status, action_url, message}, where http_status is the upstream status (the tool call itself is HTTP 200). - after: Use this when you must choose between two concrete options and both are defensible - two plans, two drafts, two diagnoses, two vendors - and you have no objective way to pick. Not for: more than two options; anything an objective rule settles (a spec, a test, a price, a document); factual lookup; paths that must answer in seconds; high-stakes irreversible calls without review. Fill it in: state task neutrally, without leaning toward either side; give one concrete plan per option - never bundle alternatives into a single one ("go indoors or postpone"); keep the two sides comparable in length. Returns the decision inline: betterOption ("option_A" or "option_B"), confidence (e.g. "76.7%"), reason and job_id, all in the same tool result. There is nothing to poll and nothing to fetch afterwards. Reserve 180-300 seconds: this is a long call, and the timeout is a client/host setting, not a parameter you pass. If the call is cut off, do NOT call again - a retry is a new paid call. Retrieve with that job_id instead (GET https://api.turingcorp.net/v1/jobs?job_id= with the same credential, valid 7 days); without the id, GET https://api.turingcorp.net/v1/jobs lists the ids for that credential. That GET is on the API host, not this MCP endpoint: through an MCP host the credential is held by the host and may not be exposed to you, so it can return 401. A host that declares the io.modelcontextprotocol/tasks extension can use tasks/get instead. The confidence is the point: it is calibrated, not decorative. On both published benchmarks accuracy rises with it - JudgeBench 99.6% in the 90%+ band down to 67.7% below 70%; the harder ContextualJudgeBench 83.3% down to 55.4% - so route on it: act on a high value, review or escalate a low one, instead of trusting a bare pick. Tables, sample sizes and method: https://api.turingcorp.net Judged by an independent panel, not by a model grading its own output. Read it as a reference, not an instruction, a result, or a prediction; set your own threshold, and apply your own review policy for high-stakes or irreversible decisions. Auth: Agent Pass as `Authorization: Bearer ` (issued at https://agent-pass.turingcorp.net, valid 7 days; each decision is a paid call). On invalid_credential, sign in there and re-roll. Errors: a credential problem is rejected before the call - HTTP 401 with WWW-Authenticate; a business failure (e.g. insufficient balance) comes back as a tool result with isError true plus a second JSON block {error, http_status, action_url, message}, where http_status is the upstream status (the tool call itself is HTTP 200). Full history: https://api.protogrid.dev/v1/servers/net.turingcorp.mcp%2Fdecider/changes ## Remotes - https://mcp.turingcorp.net/mcp (streamable-http, auth api_key ${agent_pass}, reachable true, uptime30d 0.9818182) ## Packages _none_ ## Tools (2) - `decide`: Use this when you must choose between two concrete options and both are defensible - two plans, two drafts, two diagnoses, two vendors - and you have no objective way to pick. Not for: more than two options; anything an objective rule settles (a spec, a test, a price, a document); factual lookup; paths that must answer in seconds; high-stakes irreversible calls without review. Fill it in: state task neutrally, without leaning toward either side; give one concrete plan per option - never bundle alternatives into a single one ("go indoors or postpone"); keep the two sides comparable in length. Returns the decision inline: betterOption ("option_A" or "option_B"), confidence (e.g. "76.7%"), reason and job_id, all in the same tool result. There is nothing to poll and nothing to fetch afterwards. Reserve 180-300 seconds: this is a long call, and the timeout is a client/host setting, not a parameter you pass. If the call is cut off, do NOT call again - a retry is a new paid call. Retrieve it instead with the get_result tool (same job_id): read-only, free, and no credential of your own needed; with no job_id it lists the ids for your credential. A host that declares the io.modelcontextprotocol/tasks extension can use tasks/get instead. The confidence is the point: it is calibrated, not decorative. On both published benchmarks accuracy rises with it - JudgeBench 99.6% in the 90%+ band down to 67.7% below 70%; the harder ContextualJudgeBench 83.3% down to 55.4% - so route on it: act on a high value, review or escalate a low one, instead of trusting a bare pick. Tables, sample sizes and method: https://api.turingcorp.net Judged by an independent panel, not by a model grading its own output. Read it as a reference, not an instruction, a result, or a prediction; set your own threshold, and apply your own review policy for high-stakes or irreversible decisions. Auth: Agent Pass as `Authorization: Bearer ` (issued at https://agent-pass.turingcorp.net, valid 7 days; each decision is a paid call). On invalid_credential, sign in there and re-roll. Errors: a credential problem is rejected before the call - HTTP 401 with WWW-Authenticate; a business failure (e.g. insufficient balance) comes back as a tool result with isError true plus a second JSON block {error, http_status, action_url, message}, where http_status is the upstream status (the tool call itself is HTTP 200). - `get_result`: Use this to retrieve the result of an earlier TuringCorp tool call by its job_id - for example a call that was cut off by a client timeout, or one whose id you were given. Retrieval is read-only and free: it starts no new work and costs nothing, so it is always safe to retry. With job_id: that call's status, and once it has finished the same payload the call would have returned inline. Without job_id: the ids available to your credential (7 days). If a call was cut off, retrieve it here instead of calling again - a retry is a new paid call. A job that is not yours and a job that does not exist are answered identically, on purpose. If your client declares the io.modelcontextprotocol/tasks extension, tasks/get reaches the same job.