docs Human onboarding
Explain models, datasets, jobs, streaming behavior, limits, and operational workflows in a portal designed for reading and exploration.
Teach developers and agents the same current API.
AI and data products evolve quickly, and their users work across notebooks, terminals, applications, and coding agents. Doctorine keeps the portal, verified code, machine context, and change workflow connected to the release that made them true.
One API can meet users in three modes.
Reading, coding, and delegating to an agent are different interactions. They can still inherit the same terminology, examples, operations, and release identity.
sdk · cli Programmatic use
Deliver verified TypeScript, Python, and Go SDKs plus CLI workflows for local development and CI without separating them from the source release.
mcp · skill Agent discovery and action
Expose public read-only context through llms.txt and portal MCP, then use scoped workspace MCP and Agent Skills for approved workflows.
Keep discovery open and change control explicit.
A public portal MCP server helps an agent read published documentation. The authenticated workspace MCP uses OAuth, project scope, permissions, and a staged change model for authoring workflows.
Explore MCP access ↗- Public read-only discovery does not grant workspace mutation.
- Workspace changes are prepared on a branch and remain reviewable.
- Agent Skills package stable instructions and references as open files.
- CLI JSON output and stable exit codes support deterministic automation.
Move quickly without teaching from stale context.
The release model connects authored explanation, generated artifacts, public machine routes, and agent instructions. A model or client can change quickly; the evidence behind what you publish remains inspectable.
Make one complex workflow easy to discover and run.
Choose the path developers repeat most, then deliver its human, code, and agent surfaces together.