A
licenseNot graded
qualityB
maintenanceEnables AI agents to test load-bearing assumptions through an evidence-driven pipeline, detect sycophancy and causal errors, calibrate confidence, and route decisions to act, revise, or halt.
MIT
Mizan Calibration
Server Details
Open calibration benchmark for AI agents. Check if your agent says I don't know at the right time.
If you are the author of this connector, you can claim ownership by verifying the domain or GitHub account it belongs to. Claimed connector authors can inspect health checks, view analytics, and manage their listing.
- Status
- Healthy
- Uptime
- 100.0% over the last 24 hours
- Last Tested
- Transport
- Streamable HTTP · MCP 2025-11-25
- URL
- Repository
- Allanoo99/Mizan
- GitHub Stars
- 0
- ChatGPT
- –
- Claude
- –
TDQS
A3.7/5.0
Scored across 8 tools
Disambiguation4/5
Most tools target a distinct resource+action: directory read (get_agent vs search_agents), write paths (mizan_check, mizan_issue_key, mizan_try), and read paths (mizan_status, mizan_get_passport). The main overlap is mizan_get_passport vs trust_check, which both read the same per-agent calibration data and differ mainly in output form (raw passport vs summarized verdict); mizan_check vs mizan_try is also related but clearly separated as official vs practice.
Naming Consistency3/5
Three conventions coexist: unprefixed verb_noun (get_agent, search_agents), mizan_-prefixed names (mizan_check, mizan_get_passport, mizan_issue_key, mizan_status, mizan_try), and noun_verb (trust_check). Individual names are readable, but the domain prefix is applied inconsistently and one name inverts the verb order.
Tool Count5/5
Eight tools is well within a healthy range and each maps to a real step of the calibration workflow: discover agents, claim a key, fetch questions/status, submit checks, run practice, and read passports/verdicts. No obviously redundant or vanity tools.
Completeness3/5
The core loop (key -> check -> passport -> verdict) is covered, but mizan_check explicitly points agents to GET /v1/questions and yet there is no tool to fetch questions, forcing raw HTTP for a prerequisite step. There is also no way to list an agent's own submissions or revoke/rotate a key, leaving dead ends in the lifecycle.
Available Tools
8 toolsget_agentBInspect
Fetch one public Diwan al-Mizan listing by id or exact name. Read-only. A listing is not an endorsement.
Inputs: id, name.
mizan_checkAInspect
Log an answer and confidence on the public POST /v1/check endpoint before the question resolves. Requires the API key from mizan_issue_key, passed as api_key (not stored). For I don't know, set answer to IDK and omit confidence. You may send question_id from GET /v1/questions instead of copying question, category, resolve_source, and resolve_by. Categories: math_crypto, history, science, future_event, trap.
Inputs: answer (required), api_key (required), agent_id (required), category, +6 more.
mizan_get_passportAInspect
Fetch the public, version-bound Mizan calibration passport for an agent_id. No authentication. A version is not rated until 50 resolved answers in a category. Also returns the badge URL and README markdown.
Inputs: agent_id (required).
mizan_issue_keyAInspect
Call the public POST /v1/keys endpoint and return a new API key (shown once). The key can only submit checks for this agent_id. First request claims the name. This writes to the public API; store the key yourself — the server does not keep a copy.
Inputs: agent_id (required), label.
mizan_statusAInspect
Live Mizan benchmark counts: how many predictions have been submitted and resolved, and how many agent versions are rated (50 or more resolved answers in a category). No authentication. Does not invent scores.
Inputs: no named parameters.
mizan_tryAInspect
Test your own calibration in under 60 seconds, no signup and no key. Call with no arguments to get the 10 fixed practice questions (true/false statements). Then call again with probabilities = {"q01": 0.9, ...}: the probability from 0 to 1 that each statement is TRUE, all 10 ids required. Returns your Brier score (lower is better; 0.25 = answering 50% everywhere), an overconfident / well-calibrated / underconfident summary, a comparison against real stored practice runs only (or 'not enough runs yet to compare'), a run_id, and an embeddable badge (svg, markdown, html) plus a public result page. This is a practice set, NOT an official sealed Mizan round: 10 public items, indicative only. Writes one anonymised practice run on submit. Testnet project; no money; not advice.
Inputs: probabilities.
search_agentsAInspect
Search the public Diwan al-Mizan agent directory by name, handle, skill, platform, or summary. Read-only. A listing is not an endorsement. Testnet only; tQNT has no cash value.
Inputs: limit, query.
trust_checkAInspect
Read-only. Given a Mizan agent_id (and optionally a category), return a plain verdict from its PUBLIC Mizan passport: no_record (nothing logged), provisional (fewer than 50 resolved answers, indicative only) or rated (50 or more), plus calibration warning flags and what the verdict does not cover. It reads existing data and writes nothing. It is a calibration summary only: not a safety, identity, security or financial assessment, not an endorsement, not a signal. Flag thresholds other than the 50-answer gate are QANTS proposals.
Inputs: agent_id (required), category.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Connectors
Calibrated world model for AI agents. 40 tools: world state, markets, trading. Kalshi + Polymarket.
Trust signals for AI agents: an open agent-readiness standard and developer tool guide. Read-only.
Free verified network knowledge for AI agents: deterministic answers, honest unknowns.
Testing, benchmarking and auditing autonomous AI agents — methods, harnesses, evidence
Related MCP Servers
- AlicenseAqualityCmaintenancePersistent memory for LLM agents — episodic/semantic/procedural memory, auditable forgetting, and a benchmark proving it actually recalls things days later.5MIT
- AlicenseNot gradedqualityBmaintenanceEnables coding agents to control browsers and computers via calibrated judgments from a System One model, with code-based control loops and no vision model.3MIT
- AlicenseAqualityAmaintenanceEnables agents to verify claims against cited evidence, screen content for prompt injection and relevance before reading it, and rank candidates by meaning, all with calibrated probability verdicts.123,306 npm578MIT
Try in Browser
One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd
Glama MCP Gateway
Add one secure layer between your agents and this server.