Practical overview
When it fits
Use SynapsCLI when agent work belongs in a terminal workflow, when a session should outlive a terminal window, or when another program needs to communicate with an agent engine over JSON-RPC.
Who it is for
- Developers who want to interact with AI agents from a terminal.
- Users coordinating named agents in parallel or keeping sessions alive across terminal windows.
- Developers integrating agent sessions into other programs through JSON-RPC or adding extensions.
Documented uses
- Launch an interactive model-backed agent session from a terminal.
- Delegate work to named subagents running in parallel.
- Continue or share a live session across terminal clients using the daemon.
- Drive the agent engine from another program through JSON-RPC over stdio.
- Extend the runtime with plugins and MCP servers.
Requirements & limits
The README documents Linux and macOS support and says Windows is not supported. It provides a Cargo installation route. Apache-2.0 is the repository license.
- The runtime supports hosted providers and local model runtimes; provider access has upstream account and configuration dependencies.
- Local Ollama is targeted at
http://localhost:11434/v1by default; another endpoint can be set in configuration or withLOCAL_ENDPOINT. - Azure OpenAI, Amazon Bedrock and Google Vertex AI routes are documented as text-only; modes requiring tools are unsupported on these routes.
- The configuration file can set model, thinking level, theme, identity prompt, disabled tools and provider settings.
- Plugins run as separate processes using JSON-RPC over stdio, with permissions declared in a manifest.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Install and launch SynapsCLI
Install the CLI through Cargo, authenticate or prepare a local model, and open the terminal interface.
Before you start
- Linux or macOS
- Rust and Cargo for the documented Cargo installation
- A supported provider login, or a running local model such as Ollama
Steps
Install with the command below.
cargo install synapsRun
synaps loginand choose a provider, or start Ollama locally if using a local model.Run
synapsto launch the terminal UI.
Expected outcome
The terminal interface opens for working with a selected model.
Guide limitations
- Windows is not supported.
- Provider access can depend on upstream account, entitlement, registration, policy, region and quota requirements.
- Google Gemini OAuth requires a Synaps-owned Google Desktop OAuth client registration.
Add a JSON-RPC extension
Prepare a plugin folder and manifest so SynapsCLI can load a separate-process extension on its next boot.
Before you start
- A plugin implementation in any language that can communicate using JSON-RPC 2.0 over standard input and output
- A plugin manifest declaring hooks, permissions and keybindings
Steps
Create a plugin folder under
~/.synaps-cli/plugins/.Add
.synaps-plugin/plugin.jsonwith the extension manifest.Place the implementation in the folder; the documentation gives Python and JavaScript as examples.
Restart SynapsCLI so the extension is live on next boot.
Expected outcome
SynapsCLI can load the extension as a separate process, with access governed by its permissions manifest.
Guide limitations
- The documented setup does not specify extension behavior or guarantee compatibility for a particular plugin.
- Extensions become live on next boot, not immediately after creating the files.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 13 October 2026
- Original website title
- HaseebKhalid1507/SynapsCLI · Official homepage title · retained in its source language

