Cruxible Core
Give agents a YAML-defined, graph-backed decision engine that returns deterministic answers with receipts instead of guessed policy.
Overview
Cruxible Core is an MCP server for the Build phase that runs YAML-defined rules on a queryable graph and returns deterministic decisions with proof receipts.
What is this MCP server?
- Define rules in YAML and evaluate them through a graph query surface
- Returns deterministic decisions with receipts for auditability
- CRUXIBLE_MODE env: admin (default), graph_write, or read_only
- CRUXIBLE_ALLOWED_ROOTS restricts cruxible_init to approved absolute paths
- PyPI package cruxible-core v0.1.2 with stdio transport
- Package version 0.1.2 on PyPI identifier cruxible-core
- Three documented permission modes via CRUXIBLE_MODE: admin, graph_write, read_only
Community signal: 6 GitHub stars.
What problem does it solve?
Agent workflows need consistent yes-or-no policy results, but plain LLM reasoning cannot show which rules fired or guarantee the same input always maps to the same output.
Who is it for?
Builders designing agent guardrails, internal policy bots, or compliance-adjacent automations who want proof trails.
Skip if: Quick one-off scripts that only need a single API call without maintained rule YAML or graph modeling.
What do I get? / Deliverables
You expose cruxible-core over MCP so agents get graph-grounded, receipt-backed decisions you can log and replay.
- Running stdio MCP server exposing graph query and decision tools
- Receipt-backed deterministic answers agents can cite in workflows
Recommended MCP Servers
Journey fit
Rule graphs and provable decisions matter most while building agent systems, but the same engine supports ship-time review and operate-time policy checks. Agent-tooling is where you wire deterministic guardrails and queryable knowledge graphs into MCP workflows.
How it compares
Deterministic rules-plus-graph MCP engine, not a general chat memory or vector RAG plugin.
Common Questions / FAQ
Who is Cruxible Core for?
Solo developers building agent products who need repeatable decisions documented with receipts rather than free-form model answers.
When should I use Cruxible Core?
Use it while building or shipping agent tooling whenever policy, routing, or eligibility must stay consistent across runs and tools.
How do I add Cruxible Core to my agent?
Install cruxible-core from PyPI, set CRUXIBLE_MODE and optional CRUXIBLE_ALLOWED_ROOTS, register the stdio MCP server, then point agents at your YAML rules and graph per cruxible-ai/cruxible-core docs.