Support ForgeCode as an official client (currently rejected with "unauthorized client detected")

[chindris-mihai-alexandru]

Hi AgentRouter team,

I'd like to request official support for ForgeCode as a client.

Why this matters

ForgeCode is an open-source terminal coding agent with first-class multi-provider support:

• https://github.com/tailcallhq/forgecode

AgentRouter already documents official integrations for:

• Claude Code
• Codex
• Gemini CLI
• RooCode
• Kilocode
• Qwen Code
• Droid CLI

I would like to add ForgeCode to that list as well.

Current behavior

Direct/generic API access from ForgeCode-style requests is rejected with:

unauthorized client detected, contact support for assistance at https://discord.com/invite/V6kaP6Rg44

I reproduced this against:

• GET https://agentrouter.org/v1/models
• POST https://agentrouter.org/v1/chat/completions
• POST https://agentrouter.org/v1/messages

Important detail

The same token does get past client authentication when used with the official Claude Code client configured as documented with:

• ANTHROPIC_BASE_URL=https://agentrouter.org/
• ANTHROPIC_AUTH_TOKEN=...
• ANTHROPIC_API_KEY=...

With Claude Code, I get quota/model-entitlement errors instead of unauthorized client detected, which strongly suggests:

• the token is valid
• the current block is client-specific / allow-list based

What I have ready on the ForgeCode side

I already prepared a local ForgeCode patch that supports AgentRouter with a single token and official route families:

• OpenAI-compatible route:
  • https://agentrouter.org/v1/chat/completions
• Anthropic-compatible Claude route:
  • https://agentrouter.org/v1/messages
• model listing:
  • https://agentrouter.org/v1/models

The local patch is tested on the ForgeCode side, but real usage is blocked by the current client rejection.

Request

Please either:

1. officially support / whitelist ForgeCode as a client, or
2. document the required headers / client requirements if ForgeCode support is intended

If support is enabled, I am happy to follow up with:

• a ForgeCode upstream PR for first-class AgentRouter support
• a docs PR to agentrouter-org/docs adding ForgeCode setup instructions

Related external report

Another project hit the same AgentRouter error:

• Agentrouter not working anomalyco/opencode#5060

Thanks.

Co-Authored-By: ForgeCode noreply@forgecode.dev

[chindris-mihai-alexandru]

Follow-up after broader live client validation:

• ForgeCode / generic requests are still blocked by AgentRouter with unauthorized client detected on:
  • /v1/models
  • /v1/chat/completions
  • /v1/messages
• Claude Code is now verified to work end-to-end with the same AgentRouter token on claude-opus-4-6.
• Codex is currently not viable because modern Codex requires /v1/responses, which AgentRouter does not expose; I opened Codex docs are stale: current Codex no longer supports wire_api = "chat" #22 for the stale/broken docs.
• Qwen Code reaches AgentRouter but crashes after a Claude response with Cannot read properties of null (reading 'choices'); I opened Qwen Code docs need clarification: documented model setup does not cleanly work in live testing #23.
• Droid CLI reaches first token, then crashes with null is not an object (evaluating 'H.usage'); I opened Droid guide uses legacy config format and current CLI crashes with claude-opus-4-6 #24 and also filed Factory-AI/factory#987.

So at this point the live picture looks like:

• Claude Code: works
• Codex: docs stale / currently broken
• Qwen Code: partial
• Droid CLI: partial
• ForgeCode / generic clients: blocked with unauthorized client detected

That still points to ForgeCode needing explicit AgentRouter-side client support / whitelisting before a first-class ForgeCode provider can be upstreamed confidently.

Minor Claude Code note: in my working setup, keeping only ANTHROPIC_API_KEY was sufficient; persistent ANTHROPIC_AUTH_TOKEN produced auth-conflict warnings and was not needed for the working AgentRouter configuration.

Co-Authored-By: ForgeCode noreply@forgecode.dev