goose Roadmap (October 2025)

[DOsinga]

🪶 goose OSS Roadmap (October 2025)

Welcome to the latest roadmap for Goose OSS! This document outlines our current focus areas, how we’re improving the platform, and what’s next for the Goose community.

⸻

🛠️ Improve Development Process

Both open issues and pull requests have gone down by about 30% in the last three months. We want to further improve this and aim for around 50 open PRs on average — given we handle about 100 per week, that seems reasonable.

Our new SLA targets:
• 📉 Keep open issues under 250
• 📬 Keep open pull requests under 70
• ⚡ Respond to new issues and PRs within two business days
• ⏱️ Average PR lifetime: < 7 days
• 🏷️ All new issues labeled within 3 days (bug, feature, priority, or needs clarification)

We’ll also optimize our CI setup for more automated testing and faster run times.

⸻

🧩 Better Failure Handling

Goose has become more stable over the past few months, but managing the matrix of n providers (each with multiple models) and m different MCP servers remains complex.

We plan to:
• ✅ Improve testing across providers
• 🧠 Enhance how errors bubble up from MCP servers and providers
• 💡 Add clear recovery hints for users
• 📤 Make it easier to share diagnostics with the team
• 🪶 Enable Goose itself to interpret and respond to errors intelligently

⸻

🧪 Consistent Testing for Top Providers

We’ll ensure uniform test coverage across the most used providers, verifying that all core features behave consistently. This includes automated regression testing and provider-specific diagnostics to catch integration issues early.

⸻

💻 Improve the Installation Process

A significant number of reported issues stem from installation problems across different operating systems. Our current process is somewhat ad hoc, and we’ll address that by:
• 🧭 Defining the best installation method per platform for both the desktop app and CLI
• ⚙️ Simplifying dependency management
• 🚫 Reducing (though not fully eliminating) “dependency hell”

This should make onboarding smoother and reduce friction for new users.

⸻

🦢 Turn Goose(d) into an API Server

goosed — the executable behind the desktop app — is already a REST-based API server at its core.

We’ll:
• 📘 Standardize the API
• 🔢 Move to versioned endpoints
• 🧱 Refactor internals to support multiple concurrent agents

Once the API is published, anyone will be able to build their own Goose clients, including Goose itself.

⸻

🧠 Adopt the Latest MCP Implementation

We’ve made strong progress implementing MCP, but a new version will be announced in November. We’ll stay up to date by:
• 🔄 Adopting the latest MCP release
• 🧮 Adding sampling and elicitation support

This ensures Goose remains fully compatible with the evolving Model Context Protocol ecosystem.