The wire your
AI stack
runs on.
One endpoint. Any model. Full observability.
No configuration creep.
Get your key
Free. No credit card. Get an API key in seconds.
Route any model
One endpoint routes your prompt to Claude, GPT-4o, or Gemini. Automatic fallback. No cold-starts.
Observe everything
Every call is traced. Tokens, latency, model used. Inspect in the dashboard or pull via API.
Unified Router
One endpoint. Claude, GPT-4o, Gemini. Auto-routing by cost, latency, or capability.
Tool Wiring
Register any HTTP endpoint as a tool. Schema-validated. Hot-reloadable.
Observability
Every call traced end-to-end. Tokens, latency, model, tool calls, errors.
Hot-swap
Swap models mid-session without breaking context or tool bindings.
skill.md
Machine-readable skill spec at /skill.md. Drop it into any agent.
OpenAI Compatible
Drop-in replacement for any OpenAI client. Change one URL.
One URL change.
curl -X POST https://filament.works/api/route \
-H "Authorization: Bearer fl-a3f9..." \
-H "Content-Type: application/json" \
-d '{"model":"auto","prompt":"..."}'{
"id": "tr_01j8xk4m...",
"model": "gemini-1.5-pro",
"choices": [{
"message": {
"role": "assistant",
"content": "An embedding is..."
}
}],
"usage": {
"prompt_tokens": 18,
"completion_tokens": 94
},
"filament": {
"model_used": "gemini",
"latency_ms": 287,
"status": "ok"
}
}Every call traced.
| timestamp | session | model | tokens_in | tokens_out | latency | status |
|---|---|---|---|---|---|---|
| 14:32:01 | sess_k9x2 | gemini-1.5-pro | 18 | 94 | 287ms | ok |
| 14:31:58 | sess_k9x2 | claude-sonnet | 412 | 638 | 891ms | ok |
| 14:31:45 | sess_m4r1 | gpt-4o | 67 | 201 | 423ms | ok |
| 14:31:22 | sess_m4r1 | claude-sonnet | 89 | 145 | 1102ms | fallback |
| 14:30:59 | sess_z8c3 | gpt-4o | 230 | 418 | 389ms | ok |
Full trace logs in your dashboard. Or pull via API.