BRYDGE
Checks agent work against the destination system's own records, not the agent's word
- Pricing
- Paid
- Platforms
- CLI, API
- Runs
- Locally
- Version
- 0.2.0
- Listing updated
- 2 days ago
About BRYDGE
BRYDGE is a paid MCP server for security. It runs locally on your machine and is started by your AI client.
The approach has two halves. Before acting, an agent calls brydge_supervise and gets ALLOWED, with an authorization id to write into the record the action creates, or ESCALATED, meaning a person decides and the agent must not proceed. Afterwards BRYDGE reads the destination's records, such as a payment processor, with its own read-only credential and says whether the work happened as permitted. What the agent reported is kept alongside and never decides the finding.
The agent identity comes from the BRYDGE_ACTOR setting, never from the model, so one agent cannot borrow another's permissions. Setup in BRYDGE comes first: issue an API key, declare what the action is worth, register a destination (a preset exists for Stripe refunds) and issue a mandate. A new agent starts with room for one action a day, and verified-true reports raise that.
The package is brydge-mcp on npm and runs over stdio. BRYDGE charges a share of the declared value of each action it checks.
Install
Run in a terminal
claude mcp add brydge --env BRYDGE_API_KEY='<BRYDGE_API_KEY>' --env BRYDGE_ACTOR='<BRYDGE_ACTOR>' -- npx -y brydge-mcpAdd to claude_desktop_config.json
{
"mcpServers": {
"brydge": {
"command": "npx",
"args": [
"-y",
"brydge-mcp"
],
"env": {
"BRYDGE_API_KEY": "<BRYDGE_API_KEY>",
"BRYDGE_ACTOR": "<BRYDGE_ACTOR>"
}
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"brydge": {
"command": "npx",
"args": [
"-y",
"brydge-mcp"
],
"env": {
"BRYDGE_API_KEY": "<BRYDGE_API_KEY>",
"BRYDGE_ACTOR": "<BRYDGE_ACTOR>"
}
}
}
}Add to .vscode/mcp.json
{
"servers": {
"brydge": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"brydge-mcp"
],
"env": {
"BRYDGE_API_KEY": "<BRYDGE_API_KEY>",
"BRYDGE_ACTOR": "<BRYDGE_ACTOR>"
}
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"brydge": {
"command": "npx",
"args": [
"-y",
"brydge-mcp"
],
"env": {
"BRYDGE_API_KEY": "<BRYDGE_API_KEY>",
"BRYDGE_ACTOR": "<BRYDGE_ACTOR>"
}
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"brydge": {
"command": "npx",
"args": [
"-y",
"brydge-mcp"
],
"env": {
"BRYDGE_API_KEY": "<BRYDGE_API_KEY>",
"BRYDGE_ACTOR": "<BRYDGE_ACTOR>"
}
}
}
}Configuration
BRYDGE_API_KEYYour BRYDGE API key. It starts with brydge_sk_; issue one in BRYDGE on the Connect page.RequiredBRYDGE_ACTORThe name BRYDGE knows this agent by, such as agent:refund-ops. Mandates are issued to this name.RequiredBRYDGE_URLWhere BRYDGE runs. Leave it unset for BRYDGE's hosted service.
Tools this server provides (5)
Taken from the project documentation; the running server may differ.
- brydge_supervise
- Ask for permission before acting
- brydge_report_outcome
- Report what the agent believes happened
- brydge_verify
- Read the destination's records and say whether the work happened
- brydge_get_finding
- Return findings so far
- brydge_get_headroom
- Show how many actions the agent may take in 24 hours
What it does
- Ask before acting
- brydge_supervise returns ALLOWED with an authorization id or ESCALATED to a person.
- Verify outcomes
- brydge_verify reads the destination's records and says whether the work happened.
- Daily headroom
- brydge_get_headroom shows how many actions are allowed in 24 hours before a person is asked.
- Stored findings
- brydge_get_finding returns results so far without rereading records.
- Fixed agent identity
- The agent name is set in configuration, not chosen by the model.
Good for
- Confirm an agent's refund really appeared in Stripe
- Require human approval for actions beyond an agent's mandate
- Audit whether agent-reported work matches system records
- Limit how many actions a new agent may take per day
Pricing
Paid
Charges a share of each action's declared value; rates not published
Prices change. Check brydge-ai.com before you buy.
BRYDGE FAQ
What does it need to run?
A BRYDGE API key and an actor name, set as BRYDGE_API_KEY and BRYDGE_ACTOR.
What happens without a mandate?
Every call goes to a person, and the agent acts only once they allow it.
Does the agent's report decide the result?
No. The report is kept beside the finding but never decides it.
How is it priced?
BRYDGE charges a share of the value declared for each action; exact rates are not given in the README.
Is BRYDGE free?
No. BRYDGE is a paid product.
How do I add BRYDGE to Claude Code?
Run this command in your terminal: claude mcp add brydge --env BRYDGE_API_KEY='<BRYDGE_API_KEY>' --env BRYDGE_ACTOR='<BRYDGE_ACTOR>' -- npx -y brydge-mcp
What are the best alternatives to BRYDGE?
The closest alternatives to BRYDGE are REA, Claude Code Ultimate Guide, Agent Roast, Midplane and Leash.
Alternatives to BRYDGE

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