Any model, no lock-in

The autonomous coding agent that works with any model

Claude Code-level autonomy — explore, plan, execute, verify — with the freedom to use Claude, Gemini, DeepSeek, or any OpenAI/Anthropic-compatible API. VS Code, desktop, and CLI.

Star on GitHub · Free forever · Bring your own key

Spryv — autonomous coding agent
Spryv desktop app — two-column layout with agent chat on the left and live state tree on the right

No subscription. Bring your own key.

Claude Gemini DeepSeek Kimi K2 OpenRouter Vertex AI any private endpoint
Interfaces

One agent. Three surfaces.

VS Code extension, standalone desktop app, and CLI. Same engine, same power — choose how you work.

Spryv VS Code chat panel showing an autonomous auth middleware refactor with thinking blocks, tool calls, and a success result

VS Code Extension

Chat panel with live thinking blocks, tool call results, token tracking, and session history — right inside your editor.

v1.2.46 · Requires VS Code 1.85+
Capabilities

Everything Claude Code can do — with any model

A complete autonomous agent loop. Not a chat wrapper, not a copilot. A system that explores, plans, implements, reviews, and verifies — end to end.

01

Full Agent Loop

Explore → Plan → Execute → Verify. Reads your codebase, maps the call graph, writes an explicit plan, applies changes through a built-in architect review, runs the tests, reports back.

02

Any Model — No Lock-In

Works with any OpenAI or Anthropic-compatible API — Claude, Gemini, DeepSeek, Kimi K2, OpenRouter, Vertex AI, or a private self-hosted endpoint. One flag to switch.

03

Sub-Agent Architecture

Spawn parallel specialized sub-agents, each with its own context and tool set. Define custom agents in a Markdown file — no code required.

04

Live State Tree

Confirmed facts, pending tasks, file changes, sub-agent status — updated every iteration. A health-check monitor detects loops and stalls and self-corrects.

05

Custom Skills & MCP

Define /slash-command skills in a Markdown file and invoke them by name in any interface. Connect any MCP server for additional tools.

06

Persistent Sessions

Every session saved with full conversation, tool log, state snapshots, and cost breakdown. Resume or fork any session exactly where you left off.

How it works

Give it a task. Watch it ship.

Spryv runs a structured loop modeled on how a senior engineer actually works — not a single prompt, but a full autonomous process.

  1. 1

    Give it a task

    Describe what you want in plain English. Works with whichever model you have a key for — Claude, Gemini, DeepSeek, or any compatible API.

  2. 2

    Explores, plans, and reviews

    Reads the relevant files, maps the call graph, and writes an explicit plan before touching anything. Every patch goes through a multi-round architect review before being applied.

  3. 3

    Executes and verifies

    Implements the changes, runs the test suite, and reports cost. A health-check monitor detects loops and self-corrects. You get a clean diff and a summary of what changed.

Get Spryv

Download & install

Free to use, for individuals and teams. Bring your own API key.

VS Code Extension

v1.2.46

Full chat panel with thinking blocks, tool calls, session history, and live state visualization — inside VS Code. Requires VS Code 1.85+.

Install from Marketplace One click · Auto-updates
or
ext install FaisalSalamah.spryv-vscode Download .vsix

Windows Desktop

v1.0.0

Standalone Electron app with two-column layout: chat panel + live state tree. Runs the standard Windows installer — no manual extraction. Windows 10+.

CLI

v1.2.1

Install globally and run from any directory. Phase-gated output, same underlying agent, zero UI overhead. Requires Node 18+.

$ npm install -g @spryv/agent
$ spryv "fix the memory leak in the renderer"
  • --provider to switch between Claude, Gemini, DeepSeek
  • --resume to continue any previous session
  • --chat for interactive multi-turn mode
Documentation

Everything you need to get started

From first install to defining custom agents and skills.

Contact

Get in touch

Questions, feedback, or partnership inquiries — we read everything.

[email protected]

For bugs and feature requests, use GitHub Issues.