agent-bridge-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| AGY_PATH | No | Path to the agy binary for Antigravity when not in a standard location. | |
| WINPTY_PATH | No | Path to winpty binary for Antigravity on Windows when not in a standard location. | |
| AGENT_BRIDGE_ROOT | No | Session/task state directory. Defaults to process.cwd(). | process.cwd() |
| AGENT_BRIDGE_WORKSPACE | No | Root boundary for execution targets. Defaults to process.cwd(). | process.cwd() |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| agent_sendB | Send an instruction to Claude Code or Codex CLI/SDK. Executes asynchronously in background and returns task_id. |
| agent_statusB | Check the status, output, or error of a background task dispatched via agent_send. |
| agent_sessionsC | List active or archived sessions tracked by Agent Bridge. |
| agent_cancelC | Cancel a currently running agent task. |
| agent_evidence_checkA | Run Evidence Gate validation: inspect git status/diff and execute test/typecheck/lint commands to produce empirical evidence report. |
| agent_review_loopA | Execute full autonomous development loop: Codex implements -> Evidence Gate validates -> Claude reviews -> Auto-revision (up to max 2 revisions limit). |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 6 tools
Each tool targets a distinct action: send, cancel, check evidence, check status, list sessions, and run a full review loop. The only minor overlap is between agent_send and agent_review_loop, as both can dispatch work, but the latter is a higher-level workflow clearly distinguished by name and description.
All tool names consistently follow the agent_verb or agent_verb_noun pattern (e.g., agent_send, agent_evidence_check, agent_review_loop). The naming is predictable and readable across the set.
With 6 tools, the surface is appropriately scoped for an agent orchestration and review bridge. Each tool provides a distinct function without overwhelming users or leaving the set feeling thin.
The set covers the core workflow: dispatch, status, cancel, and an integrated review loop. Evidence check and session listing add useful supporting operations. A minor gap is the lack of an explicit tool to modify or configure review loop parameters, but the overall surface is solid.