Fulcrum
Enforce AI usage policies, cap spend, and observe agent calls through Fulcrum’s governance MCP while you ship agent features.
Overview
Fulcrum is an MCP server for the Operate phase that connects agents to AI governance for policy enforcement, cost control, and observability.
What is this MCP server?
- MCP server for AI governance: policy enforcement, cost control, and observability
- stdio transport via npm @fulcrum-governance/sdk (0.1.1) and PyPI fulcrum-governance (0.1.1)
- Environment variables: FULCRUM_API_ENDPOINT, FULCRUM_TENANT_ID, optional FULCRUM_API_KEY
- gRPC-oriented Fulcrum API endpoint configuration for tenant-scoped governance
- Server package version 1.1.0 in schema pointing at Fulcrum-Governance/fulcrum-io
- Two stdio packages listed at version 0.1.1 (npm and PyPI)
- Server schema version 1.1.0
- Three documented environment variables on npm package (endpoint, tenant, optional API key)
What problem does it solve?
Your agent workflows have no tenant-aware guardrails, so tool misuse and runaway model spend are invisible until the bill or an incident lands.
Who is it for?
Indie builders running multi-tenant or billing-sensitive agent features who already deploy or trial Fulcrum governance infrastructure.
Skip if: Hobby prompts-only projects with no Fulcrum backend, or teams that only need a static linter skill with no runtime policy engine.
What do I get? / Deliverables
With Fulcrum wired over MCP, agents can consult governance policies and telemetry through a single gRPC-backed control plane you configure per tenant.
- MCP-mediated governance checks and telemetry against your Fulcrum tenant
- Documented env-based wiring for npm and PyPI stdio transports at package version 0.1.1
Recommended MCP Servers
Journey fit
Governance and observability matter most once agents touch production traffic, budgets, and tenant data—classic Operate concerns for solo SaaS. Monitoring subphase captures cost and telemetry lenses; Fulcrum is cataloged here as the ongoing control plane, not a one-off code generator.
How it compares
Governance MCP bridge to Fulcrum.io, not a generic logging skill or a scraping browser server.
Common Questions / FAQ
Who is Fulcrum for?
Fulcrum is for developers and operators who use the Fulcrum governance platform and want Claude Code-style agents to enforce policies and observe AI usage via MCP.
When should I use Fulcrum?
Use it in operate and ship when agent traffic needs tenant policies, spend limits, or audit-style observability instead of unconstrained tool access.
How do I add Fulcrum to my agent?
Install @fulcrum-governance/sdk or fulcrum-governance, set FULCRUM_API_ENDPOINT and FULCRUM_TENANT_ID (plus FULCRUM_API_KEY when required), register the stdio MCP server in your client, and restart the agent.