stagewise
An open-source agentic IDE for developers that orchestrates coding agents with full model sovereignty, supporting local, cloud, and BYOK inference providers.
At a Glance
About stagewise
stagewise is an open-source agentic IDE built by stagewise GmbH, a Y Combinator-backed company founded by Glenn Töws (CEO) and Julian Götze (CTO). The entire codebase is publicly available on GitHub under the AGPLv3 license, and the project has accumulated over 6,700 stars since its creation in April 2025. It runs locally as a desktop application and connects to any model provider — no lock-in required.
What It Is
stagewise is a next-generation agent orchestrator for software engineers. Rather than a chat-based assistant bolted onto an editor, it functions as a standalone IDE where a coding agent has full access to the developer's environment — including terminal output, browser tabs, console logs, and debugger state. Developers can run multiple agents in parallel, each tackling independent features or bug fixes, while stagewise manages context efficiency and model routing behind the scenes.
Model Sovereignty and Provider Flexibility
A core design principle of stagewise is that developers should never be locked into a single model or provider. The tool supports three distinct access paths:
- stagewise Cloud Inference: Preconfigured access to frontier and open-weight models through a stagewise account — no API keys required to get started.
- BYOK / External subscriptions: Connect existing API keys from OpenAI, Anthropic, Google Gemini, DeepSeek, Kimi, Qwen, MiniMax, Mistral, OpenRouter, AWS Bedrock, Azure, and Vertex AI.
- Local inference: Run models locally via Ollama, vLLM, LM Studio, or any provider serving the OpenAI Chat Completions API, OpenResponses API, or Anthropic Messages API. The minimum recommended context size is 150k tokens.
Agent Efficiency Architecture
stagewise claims an average cache hit rate of 87.6% by keeping the conversation prefix stable across turns. When the environment changes, the runtime sends a compact state delta rather than rebuilding the full context from scratch. The agent also automatically compresses context as tasks grow, maintaining a focused working set. This design is intended to enable long-running agentic tasks at significantly lower token cost compared to naive context rebuilding.
Enterprise Deployment Model
For teams requiring compliance and control, stagewise offers an enterprise configuration that includes regulatory and audit compliance, global inference and model configuration, access to stagewise Cloud Inference EU, SSO via OIDC and SAML, SCIM provisioning, and optional self-hosting of the stagewise Cloud. The enterprise offering is contact-sales only and documented separately at docs.stagewise.io/enterprise.
Update: stagewise 1.22.0 and Weekly Release Cadence
The project follows an active weekly release cadence, publishing "Release Week" summaries on its news page. As of July 2026, the latest release is stagewise 1.22.0 (published July 17, 2026), with release weeks documented continuously from May through June 2026. A notable announcement from May 8, 2026 declared the project's pivot to becoming "the Open Source Agentic IDE," signaling a broader platform ambition beyond its original browser-focused tooling roots. The repository shows active development with 501 forks and 76 open issues as of the latest data.
Community Discussions
Be the first to start a conversation about stagewise
Share your experience with stagewise, ask questions, or help others learn from your insights.
Pricing
Hobby
Free tier with limited access to 3 standard models and BYOK for unlimited access.
- Limited access to 3 standard models (Default, Quick, Smart)
- Bring Your Own Key (BYOK) for unlimited access and custom models
Individual
Access to all models including frontier and open-weight, with ability to extend usage via extra credits.
- Access to all models, including Frontier and Open-Weights
- Extend usage with extra credits
- Run additional models from custom endpoints
Ultra
Access to all models with 15x higher limits than Pro.
- Access to all models, including Frontier and Open-Weights
- 15x higher limits than Pro
- Extend usage with extra credits
- Run additional models from custom endpoints
Enterprise
Custom enterprise plan for teams needing compliance, control, and scale.
- Regulatory and audit compliance
- Global configuration of inference and models
- Access to stagewise Cloud Inference and stagewise Cloud Inference EU
- SSO with OIDC and SAML
- Provisioning with SCIM
- Optional self-hosting of the stagewise Cloud
Capabilities
Key Features
- Agentic coding IDE with built-in coding agent
- Multi-agent orchestration — run agents in parallel
- Full access to terminal, browser console, and debugger
- stagewise Cloud Inference with no API key required
- BYOK support for OpenAI, Anthropic, Gemini, DeepSeek, Kimi, Qwen, MiniMax, Mistral, OpenRouter, and more
- Local inference via Ollama, vLLM, LM Studio, and any OpenAI-compatible endpoint
- 87.6% average cache hit rate via stable context prefix
- Automatic context compression for long-running tasks
- IDE integration to view and apply code changes
- Reverse-engineer any website's components, style systems, and color palettes
- Temporary test changes or permanent codebase edits
- Enterprise SSO with OIDC and SAML
- SCIM provisioning
- Optional self-hosting of stagewise Cloud
- stagewise Cloud Inference EU for data residency
- Regulatory and audit compliance for enterprise
- Open source under AGPLv3
Integrations
Demo Video

