GeodesicAI
MCP server by Geodesiclabs · mcp
GeodesicAI is pre-execution governance for AI agents. Other tools do something; GeodesicAI decides whether it should happen — with deterministic, replayable verdicts, never a guess. GOVERNANCE (the core) — Validate agent outputs and authorize actions against Blueprint rules BEFORE they execute. Catches math errors, field inconsistencies, policy-limit breaches, and constraint violations. Every check returns PASS, FAIL, or REVIEW with plain-language reasons, repair suggestions, a tamper-evident determinism hash, and a proof certificate you can re-verify independently. Same input, same rules, same verdict — every time. Use authorize_execution for a go/no-go before real-world actions (payments, emails, filings, API writes); use validate for document and output checking. Requires a Blueprint (your governance contract): load_rule_pack for ready-made templates, create_blueprint for custom rules. DISCOVERY — Learn candidate rules and spot unusual records from batches of your own data, deter…
Endpoint: https://app.geodesiclabs.ai/mcp
Verification: observed. Source: crawl.
Capabilities (40)
- validate
- validate_repair
- create_blueprint
- list_blueprints
- repair
- check_blueprint_health
- compare_semantic_equivalence
- govern_inference
- get_inference_trace
- recent_inference_decisions
- verify_certificate
- profile_blueprint_robustness
- forecast
- discover_patterns
- repair_path
- counterfactual
- analyze_anomaly
- create_chain
- submit_chain_stage
- handoff_audit
- approve_rule
- reject_rule
- structural_types
- decompose_failure
- geometric_confidence