Developer quickstart

Run a governed agent from Python

The published kimss package wraps the supported agent, model, file, and vector-store paths. Its optional MCP server exposes a focused subset as tools.

1. Install and authenticate

pip install kimss

# Store your workspace key outside source control.
export KIMSS_API_KEY="kimss_..."

Kimss APIs accept X-Kimss-Key or a bearer token. Use X-Workspace-ID when your credential can address more than one workspace.

2. Use the supported agent surface

The current SDK exposes agent run and create operations, model completions, file upload, vector-store creation, and selected legacy compatibility helpers. Prefer /v1/agents/run for new agent execution.

Conversation state

Use the SDK’s conversation_id name when continuing a conversation.

Files and knowledge

Upload a file and create a vector store through the supported SDK resources.

MCP tools

The optional MCP server includes focused tools such as run agent, create agent, complete, upload, and create vector store.

3. Keep the boundary explicit

The SDK intentionally does not wrap every billing, admin, SCIM, or Backoffice route. Check the shipped SDK documentation before assuming an API family is available as a client method or MCP tool.