agent-router
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| CURSOR_API_KEY | No | Cursor Cloud Agents API key; only needed when explicitly using runtime=cloud. | |
| SLACK_BOT_TOKEN | No | Slack bot token used as an HTTP fallback if the Slack CLI is missing. | |
| AGENT_ROUTER_HQ_URL | No | Ideation HQ base URL for posting jobs to POST /hooks/jobs. | |
| CURSOR_CLOUD_REPO_URL | No | Git remote URL used for Cursor cloud launches. | |
| AGENT_ROUTER_PROJECT_ID | No | Default Ideation HQ project_id included on job posts. | |
| AGENT_ROUTER_JOBS_SECRET | No | Shared secret sent in the x-agent-router-secret header when posting jobs to Ideation HQ. | |
| AGENT_ROUTER_PROMPTS_DIR | No | Prompt catalog directory. Defaults to a sibling ../ideation-loop-system/prompts or ~/.agent-router/prompts. |
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 |
|---|---|
| list_peersA | List Agent Router peers (Cursor, Claude, Codex, Slack, GitHub, Railway, Vercel, Supabase, Stripe, Linear) and their MCP/CLI/API transports. |
| probe_peersB | Detect which peer transports are installed: CLIs on PATH, API keys in env, MCP servers in mcp.json, Slack CLI fingerprint. |
| routeA | Open the VS Code/Cursor extension for an agent peer (Claude Code, Codex/ChatGPT, Composer). Does not run claude -p or codex exec. Platform peers (github, railway, slack, …) still use CLI/API. Pass params.promptId to load a catalog prompt and params.memoryPacket to prepend shared MEMORY. |
| call_cliB | Run an allowlisted CLI for a catalog platform peer (gh, railway, vercel, supabase, stripe, slack). Rejected for claude/codex/chatgpt — open the extension instead. |
| call_apiC | Call a catalog peer HTTP API. Auth comes from the peer's authEnv (GITHUB_TOKEN, SLACK_BOT_TOKEN, CURSOR_API_KEY, ...). |
| slack_apiB | Call a Slack Web API method via Slack CLI ( |
| list_jobsB | List Agent Router cloud jobs (Cursor / Claude / Codex launches). |
| job_statusC | Get or refresh one Agent Router job by id. |
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 8 tools
Most tools have clearly distinct roles: discovery (list_peers vs probe_peers), execution by transport (call_api, call_cli, route), and job inspection (list_jobs vs job_status). The main overlap is slack_api versus call_api, since Slack is itself a catalog peer reachable via call_api; the descriptions hint at this but don't draw a crisp boundary. call_cli/route are well separated by platform-vs-agent peer scope.
All names are snake_case, but the convention mixes verb-first (call_api, call_cli, list_peers, list_jobs, probe_peers) with noun-first (slack_api, job_status) and a bare verb (route). Groups like call_* and list_* are internally consistent, but there's no single predictable pattern across the set.
Eight tools is well within the sweet spot and each maps to a distinct facet: peer discovery, transport probing, three execution channels, and job listing/status. No obvious padding or redundancy beyond the Slack/API overlap.
Discovery and execution are covered, but the job lifecycle is only half present: you can list and inspect jobs yet there is no explicit create/launch, cancel, or retry operation (route opens an extension instead of launching headlessly). Peer detail (get_peer) is also absent, so agents may hit dead ends after list_peers.