One endpoint. Any model. Full observability. Drop-in compatible with OpenAI — no configuration creep, just results.
Point your existing OpenAI calls at filament.works/api/route. Add your API key. That's it — no SDK swaps, no schema migrations, no rewriting prompts.
Set model to "auto" and filament picks the fastest available. Pin to "gpt-4o" or "claude-3-5-sonnet" for precision. Hot-swap without touching application code.
Every request logged. Latency, model used, token counts, error traces. Query from your dashboard or pull via the observe API — full visibility without extra tooling.
One endpoint dispatches to OpenAI, Anthropic, Google Gemini, and Venice. Switch models in a single field.
Every call logged with latency, token counts, model chosen, errors. Query via dashboard or REST API.
Drop-in replacement. Zero schema changes. Your existing SDK calls work unchanged.
Change your provider without touching application code. Filament resolves "auto" to the fastest available.
MCP-ready. skill.md spec. Wire tools across agents and providers through a single protocol layer.
No usage caps for core routing. Pay providers directly. Filament takes nothing from your inference budget.
Change your base URL. Add your filament key. Set model to "auto" or any specific provider. Everything else stays the same.
Filament logs latency, model selection, token usage and errors for every request. Pull logs via API or browse the dashboard — no extra instrumentation required.
View your logslive request log — last 5 calls
Free API key. OpenAI compatible. Five minutes to production.