Farhan Agent Bridge
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| FARHAN_BRIDGE_DATA | No | Select another runtime directory. By default the runtime directory is %USERPROFILE%/.farhan-agent-bridge. |
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 | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| bridge_statusA | Check local Codex and Claude subscription authentication. Returns Claude account and subscription type from official auth status. No inference call. Does not return remaining Claude allowance. |
| submit_codex_taskA | Delegate one user-authorized task to the locally signed-in Codex account using gpt-6.1-sol. Consumes that account allowance. Runs asynchronously in read-only mode with inherited user config disabled. No credentials or full ChatGPT memory are transferred; supply only relevant task context. First test: Reply exactly BRIDGE_TEST_OK without tools. The Windows computer must stay awake. Does not provide bulk scraping, Claude or Google access. |
| get_codex_taskA | Retrieve status, final answer and provider-reported token usage for a previously submitted task. Do not treat queued/running as completed. |
| submit_claude_taskA | Delegate a user-authorized text-analysis task through official Claude Code subscription login. Optional public web research with allow_web=true. No shell, local-file, MCP, messaging or publishing tools. Uses Sonnet and checks subscription auth before each task. One shared active task, 20-minute timeout, polling required. First test: Reply exactly CLAUDE_BRIDGE_TEST_OK. Never infer allowance from token counts; no API fallback is configured. |
| get_claude_taskA | Retrieve a Claude task status, final answer and usage. Cached input, cache creation and ordinary input are separate provider fields. A list-price estimate is not an actual subscription charge. Queued/running is not completion. |
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 5 tools
Each tool has a clearly distinct role: bridge authentication status, task submission, and task retrieval. The Codex and Claude variants are explicitly provider-scoped, so an agent can select the correct tool without ambiguity.
The task tools follow a consistent verb_provider_task pattern: submit_codex_task, get_codex_task, submit_claude_task, get_claude_task. bridge_status is a minor deviation because it omits the verb-style prefix, but the overall convention remains readable and predictable.
Five tools are well matched to a two-provider bridge: one shared status check plus submit/get pairs for Codex and Claude. Each tool has a clear purpose and there is no excessive surface area.
The surface covers the core lifecycle: checking authentication, submitting tasks, and retrieving status, answers, and usage for both providers. Missing cancellation or task-listing operations are minor gaps that do not block the primary bridge workflow.