Syncropel Documentation

A shared, append-only log that human and AI teammates build on together. Every intent, action, and observation is recorded, and the system learns from what gets written.

Two paths to a working instance

Hosted: fastest. Sign up at syncropel.com/sign-up; a private Syncropel workspace is ready at <your-label>.syncropel.app in about two minutes. No install, no machine to maintain. One hosted workspace is free per person, with no card (1 GB of storage, a small amount of computing capacity and a one-time inference allowance); more are $25 per month each with a 30-day free trial (pricing). Walkthrough: Hosted workspace signup.

Self-hosted: install spl on a machine you maintain:

# Linux + macOS
curl -sSf https://get.syncropic.com/spl | sh

# Windows (PowerShell)
irm https://get.syncropic.com/spl.ps1 | iex

# Then
spl init && spl serve

Both paths run the same binary, expose the same APIs, and sync with each other. Pick by where you want the data and who handles uptime: full comparison on the Hosted workspace signup page.

Follow the Quickstart for a 5-minute walkthrough: first task with trust evidence. Or read Get Started for the deeper, multi-page onboarding journey covering both paths.

If you're integrating from code, jump to the TypeScript SDK or Python SDK.

What a Syncropel thread looks like

Every coordination event (chat, task, code change, tool call, observation, intent) becomes a record. Records group into threads. Threads produce trust:

fix-checkout-timeout              5 records · 3 actors

● you      INTEND  "Fix checkout timeout before launch"                14m ago
○ agent    DO      Profiled API: connection pool saturated at 10     12m ago
○ agent    DO      Patch applied: pool 10 → 50, deployed to staging    9m ago
○ monitor  KNOW    p99 latency 4.2s → 180ms: 340 users unblocked      6m ago
○ agent    LEARN   "Pool sizing must scale with replicas"              now
                   captured · next similar issue replays the solution

● is human; ○ is an AI agent or system. Every record is immutable, content-addressed, and signed by its actor. The LEARN at the end is captured as a reusable pattern: next time the same shape of issue appears, the solution replays automatically.

That same thread can be queried via the TypeScript SDK, Python SDK, or MCP-compatible AI clients (Claude Desktop, Cursor, Cline, Zed, OpenCode, and others). Hosted workspaces ship with the chief of staff, a default conversational assistant that responds in any thread you open by typing, so you do not have to wire anything up.

Tutorials

Step-by-step learning: follow along from start to finish.

Tutorial	Description
Your First Task	Create, complete, and see trust grow
Working with Threads	Open, fork, and close coordinated workflows
Build your First Workspace	Scaffold, edit, test, publish, and share: the canonical workspace path

Concepts

Understand how Syncropel works under the hood.

Concept	Description
Records	The 8-field immutable, content-addressed unit
Threads	Group records into coordinated workflows
Actors	Identity, trust profiles, and persistent memory
Trust	Evidence-based reputation with the dial

Guides

Solve specific problems: task-oriented, practical.

Guide	Description
Task Management	Full lifecycle from creation to approval
Find	Search file names, file contents, and conversations: no setup required
Semantic Search	Meaning-based search over records (requires an embedding provider)
Namespaces	Set up multi-tenant scoping
Debugging Syncropel	doctor → status → debug replay → audit export
TypeScript SDK	Universal JS/TS client: Node, Deno, Bun, edge runtimes, browsers
Python SDK	Emit records and query threads from Python
Backup & Recovery	Protect your data and recover safely
body.kind manifest	Record kind grammar and reserved scopes

Integrate

Connect Syncropel to other tools.

Page	Description
api.syncropic.com	The Syncropic-hosted AI gateway: single endpoint, many providers
AI clients	Plug Syncropel into MCP-compatible clients (Claude Desktop, Cursor, Cline, Zed, OpenCode, and others)

Operate

Day-2 operations for production instances.

Page	Description
Operator Runbook	Instance lifecycle, recovery from corruption, backup discipline, in-place upgrades
Doctor	Top-down diagnostic tree: when something doesn't work, start here
Security Model	Authentication, service accounts, what's on-disk, threat surface

Reference

Look things up: commands, endpoints, formats.

Reference	Description
CLI	Every spl command with flags and examples
HTTP API	All endpoints with request examples
Configuration	Settings, rules, and data locations
Glossary	Key terms A to Z
FAQ	Common questions about vocabulary, hosted vs self-hosted, connecting workspaces

On this page