Platform
The platform’s functional APIs — run skills, execute workflows, orchestrate agent teams, chat, and remember. These are capabilities, not data, and they run against a live backend.
AI & Agents
- AgentsAgent definitions — roles, skills, tools, and MCP-server bindings.
- ChatChat completions (OpenAI-compatible) with sessions, artifacts, and HITL.
- Speech-to-TextTranscribe audio to text. Multipart upload; cpod-backend forwards to the governed gateway which routes to a multimodal model (e.g. google/gemini-2.5-flash).
- MemoryLong-term memory — recall, acknowledge, and notification actions.
- LLMDirect OpenAI-compatible chat completion, proxied to the gateway (not an agent loop).
Automation
Apps
- ProjectsProjects — group agents, chats, and members under a shared context.
- AppsApp file storage — upload files (with optional appId), list, search, and ask agent. APP-SCOPED FILES: pass appId on upload to scope a file to your app. List with appId to see only that app's files; without appId to see workspace/personal files.
- Knowledge basesKnowledge bases — document collections for retrieval.
- PodsPods — sandboxed compute workloads with a start/stop lifecycle.
Identity
Permissions
Runtime
Governance
Integration
- MCP ToolsOur MCP server (port 8000/mcp/) — 54 tools: memory, files, workspaces, projects, agents, search, workflows, teams, skills. HOW TO USE: 1. GET /api/v1/mcp/tools → list all tools 2. POST /api/v1/mcp/tools/{tool_id}/invoke → execute any tool 3. Or connect directly: POST /mcp/ with Bearer JWT + MCP protocol
- Model configsRegistry of LLM / embedding / reranker credentials and gateway routing.
- Feature flagsRuntime feature-flag evaluation.
- EventsEvent subscription and history for cPod resources.