Intercept — policy enforcement layer between your agent and MCP servers

[s-a-m-a-i]

If you're using mcp-use to connect agents to MCP servers, you might want guardrails on what those agents can actually do.

We built Intercept — a transparent MCP proxy that evaluates every tool call against a YAML policy file before forwarding it upstream.

Rate limits, argument validation, unconditional blocks, spend caps — all enforced at the transport layer. The agent doesn't know it's there.

delete_repository:
  rules:
    - action: "deny"
      on_deny: "Repo deletion not permitted"

create_issue:
  rules:
    - rate_limit: 5/hour
      on_deny: "Hourly issue limit reached"

Works with any MCP server. You just point your mcp-use client at Intercept instead of the server directly.

Open source, MIT licensed: https://github.com/policylayer/intercept

Site: https://intercept.policylayer.com

[aniruddhaadak80]

From my point of view, a transport-level policy proxy is the right boundary for this kind of guardrail because it works even when the agent runtime changes. The important part is that denial semantics, audit logs, and retry behavior stay explicit so the client can tell the difference between policy rejection and ordinary tool failure.

If those contracts are stable, this kind of layer can become a very practical control point for MCP deployments.