Choose your path
Who Synap is for, and where to start depending on where your pod runs, which agent you use and what data you bring.
Synap is a pod: a server that holds your records in your own PostgreSQL database, and lets the AI agents you connect read and change them under rules you set. Every change is written to an append-only event log. What is Synap? explains the model.
Who it is for
- You use an AI agent for real work (Claude Code, Codex, Cursor, ChatGPT or another MCP client) and want it to work on your records instead of starting from nothing each conversation. The context lives in your pod, so it stays when you change agents.
- You want one place for your records across domains: people and companies, tasks, notes, documents, deals, content. Each domain gets its own workspace on the same records, so a person is stored once however many workspaces use them.
- You want to see and control what agents do. Each agent has its own key and identity. You choose how much each one does on its own, from creates and edits that apply with Undo to every change asking you first; deletions and structural changes always ask you. You approve on the desktop in the Browser app (Relay, the phone app, is in beta testing).
- You run a small team. One pod holds the team's records; each person signs in as a member of the pod and connects their own agents.
- You build software and want a data store with permissions, review and an audit trail already in place: an app, a website, a script or your own agent.
Four choices
You make four choices. They are independent: any answer to one works with any answer to the others.
1. Where your pod runs
| Synap Cloud | Your own server | |
|---|---|---|
| Who operates the server | Synap | You |
| How you start | Cloud quickstart: npx @synap-core/cli init | Self-host quickstart: the installer from the backend's GitHub releases |
Both run the same code from the same releases, and the same apps and agents connect to both. Self-hosted pod or Synap Cloud lists what Synap Cloud runs for you (updates, email, billing, connected accounts) and what you take on when you run your own. A self-hosted pod can be registered with the Control Plane later; see Connect to the Control Plane.
2. Which agent works on it
Synap does not sell or meter inference: you bring your own agent. synap init connects the agents it finds on your
computer (Claude Code, Codex, Cursor), and synap connect --list lists every client it can configure. An agent you
wrote yourself gets an MCP URL and a key from synap agents add --name <name>. See Connect your AI.
3. What goes in it
- Start from a template. A workspace template sets up one domain (a CRM, a content studio, finance) with its kinds of records, fields, views and playbooks (templates in the apps). See Spaces and templates.
- Bring your data in. Import what you already have. See Import your data.
- Capture as you go. Ask your agent to save what you are working on, or capture from Relay on your phone.
4. How far you take it
- Make it yours without code: install workspaces, add kinds and fields, add capabilities and automations. See Make it yours.
- Build on it: an app on the SDK, a script on the Hub REST API, a website that reads your pod. See Build on Synap and Build a website.
If you want this, start here
| If you want to… | Start at |
|---|---|
| Try Synap with the least setup | Cloud quickstart |
| Keep your data on hardware you control | Self-host quickstart |
| Decide between the two | Self-hosted pod or Synap Cloud |
| Give your AI agent access to your pod | Connect your AI |
| Set up a domain (sales, content, finance…) in one step | Spaces and templates |
| Bring in data from other tools | Import your data |
| Track a kind of record Synap does not have | Kinds, roles and fields |
| See what you can install | Marketplace |
| Approve agent changes from your phone | Relay (in beta testing) |
| Build an app or a website on your pod | Build on Synap |
| Run your pod in production | Run |
Words you do not know are in the glossary.