Connect AI to your data pod
Give any AI — Claude Code, custom agents, LLM pipelines — direct access to your Synap data.
Who this is for
- Developers building custom apps or automations on top of Synap data
- Power users connecting Claude Code or other AI tools to their personal knowledge base
- AI builders wiring Synap as a memory + action layer for LLM pipelines
Your data pod exposes three distinct integration surfaces. Choose based on how much control you need versus how much you want to set up.
The three options
| Option | Endpoint | Best for | Pros | Cons |
|---|---|---|---|---|
| C — Skill invocation | POST /api/external/skills/:id/invoke | Calling specific named operations | Deterministic, composable, fast | Must know skill ID; no natural language |
| D — Chat stream | POST /api/external/chat/stream | Conversational AI with full pod context | Natural language, persistent history, full IS capabilities | Less predictable output; higher latency |
| E — SDK + direct API | tRPC / Hub Protocol REST | Custom apps and LLM pipelines | Full control, type-safe with @synap/sdk | Most setup required; you manage context yourself |
Option C — Skill invocation
Skills are named operations that run in a secure sandbox with full access to your pod data. Invoking them via HTTP lets any external tool — a shell script, a Claude Code tool, a cron job — trigger precise, repeatable operations.
Best when you want deterministic behavior: "extract entities from this text", "summarize overdue tasks", "create a note from these meeting notes".
Read the skill invocation guide →
Option D — Chat stream
A single endpoint backed by your pod's AI system. The AI has access to your entities, documents, skill library, and conversation history. You send a natural language query; the pod streams back a response.
Best when you want the AI to figure out what to do: "what's on my plate this week?", "draft a follow-up email based on my notes from yesterday".
Option E — SDK and direct API
For building real apps. @synap/sdk gives you typed tRPC access to every procedure on the pod. Hub Protocol REST gives you the same access surface the pod's own AI uses internally.
Best when you are building something: a custom UI, an LLM pipeline where you manage context yourself, a sync integration, or a script that reads Synap data and feeds it into a separate model.
Read the SDK and direct API guide →
API keys
All three options require an API key with the appropriate scopes.
Quick orientation
Your pod URL: https://YOUR_POD.synap.live
Option C: POST /api/external/skills/:id/invoke
Option D: POST /api/external/chat/stream
Option E: https://YOUR_POD.synap.live/trpc (tRPC)
https://YOUR_POD.synap.live/api/hub-protocol (REST)
All endpoints accept Authorization: Bearer YOUR_API_KEY. No OAuth flow required for personal access tokens.