agent command center

Turn any API into a governed, monetizable MCP endpoint.

Bouncer sits between your AI and the outside world. Turn any OpenAPI API into a governed MCP endpoint, proxy existing agents, and enforce allow/deny rules, rate limits and spend caps — without touching the agent itself.

Run a free Security Check Add your agent

free · no account needed · vet any npm package or MCP server before your agent touches it

POST https://bouncer.run/api/public/gw/my-agent

x-gateway-key: lvg_••••••••••••

{ "method": "tools/call", "params": { "name": "delete_repo" } }

403 blocked_by_policy — deny rule "delete_*"

IMPORT https://api.example.com/openapi.json

Generated 24 MCP tools

Proxy: /api/public/gw/my-api/mcp

200 tools/list ready for WebMCP

security check

Nobody reads what their agent installs. We do.

Before an npm package or an MCP server touches your machine or your agent, run the npm package security check and MCP security scanner on it: install scripts, credential access, obfuscated code and hidden tool descriptions — with a verdict you can read and a config pinned to the safe version.

Run a Security Checkfree · no account needed

$ npx -y some-hot-mcp

Turned away — 3 tool descriptions hide invisible characters

Needs a look — axios 1.20.0: runs a ‘prepare’ script on install

On the list — left-pad 1.3.0: nothing risky in it

Install scripts read as text, never executed · MCP tool lists fingerprinted

bouncer is not an llm gateway

We don't route prompts to models. We govern what your agents can touch.

LLM routers & gateways

Send your prompt to GPT, Claude or Llama
Pick the cheapest or fastest model
Count tokens and cache completions

Bouncer

Sits between your agent and its tools, APIs and MCP servers
Vets packages and MCP servers before they're installed
Enforces allow/deny lists, rate limits and spend caps on every call

Use Bouncer alongside whatever model router you already have — they answer to you, your agents answer to Bouncer.

OpenAPI → MCP in one click

Paste a Swagger or OpenAPI URL and Bouncer generates MCP tools, serves them over Streamable HTTP, and gives you a drop-in WebMCP snippet.

One proxy URL

Allow & deny lists

Whitelist the tools you trust, blacklist the rest. Wildcards supported, default-deny optional.

Keys, limits & caps

Issue scoped keys with per-minute rate limits, daily quotas and hard spend caps.

Every call logged

Tool name, decision, latency, status and tokens — for each request that passes through.

WebMCP ready

Copy a ready-to-paste WebMCP snippet from any agent so browser-based AI clients can call your governed tools directly.