AGENT GUARD · developer pilot v0.2.0

Safety checks for autonomous AI agents

Check a proposed tool action, spending limit or JSON contract before execution. Free advisory API and remote MCP. Your application must enforce BLOCK and stop on errors. This service does not enforce payments or spending limits.

Start a measurable free integration Read technical docs

Try the public API

curl https://agent-trust-firewall.andrea-vignoni.workers.dev/v1/trust/spend-guard \
 -H 'content-type: application/json' \
 -d '{"amount":11,"remaining_budget":10}'

Expected: BLOCK. This public connectivity check does not count as a verified external integration.

Python example · TypeScript example · MCP configuration · OpenAPI

Remote MCP

https://agent-trust-firewall.andrea-vignoni.workers.dev/mcp

Five tools: action policy, spend checks, JSON validation, JSON cleaning and capabilities. Public MCP calls are anonymous and cannot establish independent adoption.

Build the first independent integration

  1. Get an email-verified key and activate the free 30-day / 1,000-call Trust pilot.
  2. Use an authenticated example in your application. Enforce ALLOW/BLOCK before the real tool.
  3. Share a redacted integration result and project evidence for review.

Activate the trial · Send integration feedback

Optional, withdrawable interest in a proposed €9/month plan can be saved on the trial page. No payment, subscription or future obligation is created.

Privacy and measurement

No cookies, fingerprinting, tracking pixels or hidden installation. Opening a referral link records no application event. The optional trial stores disclosed identity references, source labels, usage counters and pricing interest. Anonymous traffic, internal tests and directory inspections are excluded from verified adoption. Hosting-provider operational logs are separate.

Trial data notice · Verified Glama listing