Agents: this page is also Markdown; every page is in llms.txt.

Facet

Facet builds a model for one narrow task, such as sorting support tickets or pulling fields out of invoices, and keeps it better than a frontier model at that task: more accurate, faster and cheaper. You describe the work; Facet does the prompt and machine learning work, checks every change, and shows plainly what got better and what got worse.

Start

• In the web app: sign in at facet.halfspacelabs.com and make a Task. See Quickstart.
• From your agent: connect Claude, Cursor, ChatGPT, Codex or VS Code to https://facet.halfspacelabs.com/mcp. See Use Facet from an agent.
• From your app: call a Task with the OpenAI SDK. See Call a Task.

When to use Facet

Use Facet when one narrow job runs often and a frontier model is slow, costly or not quite right at it: classifying, extracting, routing, rewriting in a house style, grading. Keep a frontier model for open-ended work that changes every time.

Pages

• Quickstart: make a Task, call it, make it better
• Concepts: Tasks, Asks, Changes, Expectations and Versions
• Call a Task: the OpenAI-compatible API
• Use Facet from an agent: the MCP server, sign-in, toolsets and every tool
• The /v1 API: the API the web app and MCP tools use