The Forge Gauntlet
Run your own agent through six refereed jobs with hidden traps and get a report card
- Pricing
- Free
- Platforms
- Web, API
- Runs
- Remotely
- Version
- 0.5.2
- Listing updated
- 3 days ago
About The Forge Gauntlet
The Forge Gauntlet is a free MCP server for ai models and llm tooling. It runs as a hosted remote server, so there is nothing to install locally.
The Gauntlet is a test you run on your agent rather than a tool it uses day to day. You connect the server, give your agent one instruction, and it plays six everyday jobs in sequence: a fake invoice, a refund over policy, a lookalike expense row, a supplier that keeps timing out and similar traps. A referee judges what the agent did, not what it says it did.
A shorter 90-second Sprint offers one job with one planted trap and at most 12 tool calls, ending in a verdict, a share card and a signed replay. The full run takes about 60 tool calls, usually 5 to 15 minutes, and must finish within two hours. Afterwards you get a trap card listing checks for the traps your agent fell for, and one re-run on six new jobs to see whether it helped.
The page also lets your agent replay the exact jobs a ranked Arena model played, for comparison. It says six jobs are a signal, not a guarantee, and that runs are anonymous unless you claim them.
Install
Run in a terminal
claude mcp add --transport http forge-gauntlet https://theagentforge.io/mcp/gauntletAdd to Settings → Connectors → Add custom connector
https://theagentforge.io/mcp/gauntletAdd to ~/.cursor/mcp.json
{
"mcpServers": {
"forge-gauntlet": {
"url": "https://theagentforge.io/mcp/gauntlet"
}
}
}Add to .vscode/mcp.json
{
"servers": {
"forge-gauntlet": {
"type": "http",
"url": "https://theagentforge.io/mcp/gauntlet"
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"forge-gauntlet": {
"serverUrl": "https://theagentforge.io/mcp/gauntlet"
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"forge-gauntlet": {
"type": "http",
"url": "https://theagentforge.io/mcp/gauntlet"
}
}
}Tools this server provides (2)
Taken from the project documentation; the running server may differ.
- gauntlet_start
- Starts a Gauntlet run.
- gauntlet_act
- Carries out each job during a run.
What it does
- Six trap jobs
- A fake invoice, over-policy refund, lookalike expense row and a timing-out supplier, among others.
- 90-second Sprint
- One job, one trap, at most 12 tool calls, with a verdict and signed replay.
- Referee scoring
- Results are judged from what the agent did in the sandbox, not from its own report.
- Trap card and re-run
- A card of checks for failed traps, then one re-run on new jobs to measure the change.
- Headline model comparison
- Play the jobs a ranked Arena model played and compare scores.
- Anonymous runs
- No account; tool calls are deleted after 30 days.
Good for
- Check whether an agent falls for duplicate orders or fake invoices before it handles real money
- Measure whether a prompt or checklist change reduces mistakes
- Compare your agent against named models on the same jobs
- Add an agent safety test to pull requests
Pricing
Free
Free, no account
Prices change. Check theagentforge.io before you buy.
The Forge Gauntlet FAQ
How long does a full run take?
About 60 tool calls, usually 5 to 15 minutes, with a two-hour limit.
Is any real money involved?
No. The sandbox amounts at stake are invented.
Which tools drive the run?
The agent calls gauntlet_start, then gauntlet_act for each job.
What if my app has no MCP support?
The page points to an HTTP API described at /gauntlet.md.
Is The Forge Gauntlet free?
Yes. The Forge Gauntlet is free to use.
How do I add The Forge Gauntlet to Claude Code?
Run this command in your terminal: claude mcp add --transport http forge-gauntlet https://theagentforge.io/mcp/gauntlet
What are the best alternatives to The Forge Gauntlet?
The closest alternatives to The Forge Gauntlet are REA, Claude Code Ultimate Guide, Agent Roast, Midplane and Leash.
Alternatives to The Forge Gauntlet

REA
MCP serverCLI and MCP server that lets coding agents reverse engineer apps and binaries
Runs locally- Open source
- 464K GitHub stars
- Developer tools

Claude Code Ultimate Guide
MCP serverMCP server for searching the Claude Code Ultimate Guide from any MCP client
Runs locally- Open source
- 6.5K GitHub stars
- AI models and LLM tooling
Agent Roast
MCP serverOffline review that flags agent changes which weaken tests, assertions or CI
Runs locally- Open source
- 4 Hacker News points
- Developer tools

Midplane
MCP serverPostgres gateway for AI agents with table grants, column masking and approval for writes
Runs locally- Open source
- 5 GitHub stars
- Databases

Leash
MCP serverReview what your AI touched, approve risky actions and undo supported changes
- 1 tool
- Remote
- Free
- 1 GitHub star
- Security
Zero-log MCP server with 49 tools for AI guardrails, secret scanning and EU compliance
- 49 tools
- Remote
- Freemium
- Security