One APIfor AI.
Text and code from every major provider — one API key, one balance. OpenAI-compatible: migrate by changing one base URL. Need it done end-to-end? Run agents as endpoints.
Compatible with
Price first.
Then generate.
The exact cost sits next to the button and recomputes as you change parameters — before you press it. Free starter credits on signup.
Every major model.
One key.
The fabric routes each request to the best provider — or the exact model you name. Smart modes: companyfabric/auto · fast · cheap · best · code
Text
Frontier chat and reasoning models behind the same OpenAI-compatible endpoint.
$0.3/Mtok in · Gemini 3 Flash
Code
Point Claude Code or Cursor at one base URL and swap code models per task.
$0.3/Mtok in · Qwen3 Coder Flash
Agents
Need it done end-to-end? Run agents as endpoints — same key, same balance.
Install for agentsThree doors.
One balance.
Credits
Prepay once, spend across every model. Exact metering, no subscription required.
BYOK — 0%
Bring your own provider keys and pay nothing on top. Routing and one-key ergonomics stay.
Drop-in compatible
Point any OpenAI SDK at our base URL. Same request shape, same response shape — change two lines, keep your code.
Let your agent
handle it.
Add the MCP server to Claude Code, install a skill, then just ask — “make me a 20-second product video”. Your agent plans, picks the model, generates, and reports the cost.
Install for agents# hands
claude mcp add --transport http companyfabric \
https://mcp.companyfabric.com
# know-how
/plugin install companyfabric-video-production
Already
compatible.
Keep your OpenAI SDK. Change the base URL. Every model above is now yours — with exact metered cost on every response.
import OpenAI from "openai"
const client = new OpenAI({
baseURL: "https://api.companyfabric.com/v1",
apiKey: process.env.COMPANYFABRIC_API_KEY,
})
const res = await client.chat.completions.create({
model: "companyfabric/auto",
max_tokens: 1024,
temperature: 0.7,
messages: [{ role: "user", content: "Say hello from the fabric." }],
})
console.log(res.choices[0].message.content)OpenAI-compatible
Existing SDKs and harnesses work unchanged.
Streaming
SSE with a final usage chunk carrying exact cost.
Exact metering
Every response reports what the ledger recorded.
Structured errors
Every error states its machine-actionable fix.
Every model.
One key.
Free starter credits on signup — enough to try any model in the playground before you spend a cent.