Wire agents, tools, and LLMs into deterministic pipelines — without the framework lock-in. Synapse is an open-source platform for creating, connecting, and orchestrating AI agents powered by any LLM — local or cloud. Agents use real tools: browsing the web, querying databases, executing code, reading files, managing emails, and anything else you can expose through an MCP server, a webhook, or a Python script.
SynapseOrch: AI workflows that follow the plan. Describe a job, approve the steps, run them on any model, local or cloud. MCP server included. Open source (AGPL-3.0).
Then open http://localhost:3000. See the Docker guide in the docs for custom ports and environment variable configuration.
Security: publish only the frontend port 3000 — the backend (8765) serves an internal API and should not be exposed on 0.0.0.0 (docker-compose binds it to 127.0.0.1). The container auto-generates a shared internal token so that surface is authenticated by default. For any network-reachable deployment, enable login (Settings → Security) and/or set allow_stdio_mcp=false, since registering a stdio MCP server launches local commands.
Upgrading
Install method
Upgrade command
Bash / PowerShell installer (recommended)
synapse upgrade
pip
pip install --upgrade synapse-orch-ai
npm
npm update -g synapse-orch-ai
Docker
docker pull synapseorchai/synapse-ai:latest
Upgrade note (security hardening): the backend now auto-generates a shared internal token and, under docker-compose, binds its port to 127.0.0.1 by default. The UI and its API are unaffected (the frontend proxies the backend internally). If you previously called the backend directly on :8765 (e.g. external API clients), either point them at the frontend instead — http://host:3000/api/v1|v2/... — or set SYNAPSE_BACKEND_BIND=0.0.0.0 in .env to keep publishing :8765.
Scale Mode
Run unlimited agents and orchestrations concurrently. When you need to go beyond a single process, the distributed scale layer handles the load:
ARQ worker fleet — 1 to 100+ independent workers, each running up to 20 concurrent orchestrations; autoscale with KEDA on queue depth
PgBouncer — multiplexes hundreds of worker connections into a small, stable Postgres pool
S3 artifact storage — stream large file outputs directly to AWS S3, Cloudflare R2, or MinIO
Multi-tenant quotas — per-team or per-customer concurrent run limits with HTTP 429 enforcement
Per-step checkpoint recovery — worker crashes don't lose jobs; the next worker resumes from the last completed step
Three Docker images — pull only what you need:
docker pull synapseorchai/synapse-ai:latest # full app (standalone mode)
docker pull synapseorchai/synapse-ai-api-server:latest # stateless API server (scale mode)
docker pull synapseorchai/synapse-ai-worker:latest # worker process (scale mode, run as many as needed)
The docker-compose.yml in the repo spins up the full stack. Production K8s manifests are in infra/k8s/.
Multi-Model Orchestrations — Run a different LLM at every step. Use a fast model for routing, a powerful one for analysis. You control where the compute goes.
Deterministic DAG Execution — Orchestrations follow the exact path you designed. No hallucinated detours.
Turn Anything Into a Tool — Python scripts, REST APIs, webhooks, MCP servers, or entire orchestrations — all become agent-callable tools.
Human-in-the-Loop — Pause workflows for human review. Resumable across restarts. Connect via UI, Slack, Telegram, or any messaging channel.
Scales to Millions of Requests — The distributed scale layer separates API servers, Redis job queues, and independent worker processes so you can run any number of agents or orchestrations concurrently. Start on one machine, grow to a Kubernetes cluster — the V2 API never changes.
Local-First, No Lock-In — Full local operation with Ollama. Mix local and cloud models freely. Your data stays yours.
Built-In Scheduling & Messaging — Cron-based automation with results pushed to Slack, Discord, Telegram, Teams, or WhatsApp.
14+ LLM Providers — Cloud, local, and CLI providers including Ollama, OpenAI, Anthropic, Gemini, xAI, DeepSeek, AWS Bedrock, and more.
Stable versioned REST API for building products on top of Synapse — enqueue, stream, cancel, webhooks. Docs →
Vault
Persistent file storage shared across agents and sessions. Docs →
CLI
synapse start # start backend + frontend, open browser
synapse stop # stop background processes
synapse upgrade # upgrade to the latest version
synapse uninstall # remove Synapse, wipe ~/.synapse, and uninstall the package
Roadmap
Spawn Sub-Agent Tool — Agents natively spawn and delegate tasks to temporary sub-agents mid-execution.
Compact Conversations — Automatic message history compression for large contexts.
Global Variables — Dynamic variables injectable into prompts, orchestrations, tools, and MCP environments.
Star History
Contributing
See CONTRIBUTING.md for dev setup, architecture details, how to add MCP tool servers, and the PR checklist.
License
Synapse AI is licensed under AGPL v3 — see LICENSE
S
Synapse Ai
synapseorch-ai·341 sao·
SynapseOrch: AI workflows that follow the plan. Describe a job, approve the steps, run them on any model, local or cloud. MCP server included. Open source (AGPL-3.0).
Chưa có danh sách công cụ. Máy chủ từ xa sẽ có danh sách sau lần kiểm tra tình trạng đầu tiên.
Câu hỏi thường gặp
Câu hỏi
Làm thế nào để cài đặt Synapse Ai?
Sao chép cấu hình stdio cho Cursor, Claude Desktop, Claude Code hoặc VS Code trên trang này. Cấu hình dùng bản cài đặt của nhà phát hành, không phải proxy của PlugMCP.
Synapse Ai có cần khóa API không?
Bản ghi này không liệt kê biến môi trường bắt buộc nào.
Synapse Ai dùng giao thức truyền tải nào?
Synapse Ai chạy cục bộ qua stdio.
Synapse Ai được kiểm tra lần cuối khi nào?
Kiểm tra lần cuối: 2026-11-16. SynapseOrch: AI workflows that follow the plan. Describe a job, approve the steps, run them on any model, local or cloud. MCP server included. Open source (AGPL-3.0).