# goose v1.56.0: what's new, verdict & setup Canonical: https://theagenticweekly.com/tools/goose · Released: 2026-10-23 · Last verified: 2026-10-25 · Evidence: Tested (smoke test, 2026-10-25, local Qwen2.5-3B-Instruct, not vendor models) · Vendor: aaif-goose (GitHub organization, previously block/goose) · Official site: https://github.com/aaif-goose/goose > Open-source, extensible AI agent that installs, runs, edits and tests code with the LLM of your choice. ## Verdict **Pick goose if you want an open-source agent with real subagent controls; v1.56.0 is a solid, subagent-heavy update.** Good for: - People who want an Apache-2.0 agent with a desktop app and CLI - Users who delegate work to subagents and need resume and limits - Windows desktop users who were waiting for an installer Skip if: - You want a tiny tool with few moving parts - You need a managed, supported product Tested. Tested 2026-10-25 (America/Chicago). Ran the official download_cli.sh (read first); it stopped because bzip2 is missing here, so we unpacked the same release archive with Python. "goose --version" printed 1.56.0. "goose run --no-session" with provider ollama (local model): attempt 1 called the edit tool but the text did not match and the file was unchanged; attempt 2, with exact before and after strings, printed "Edited calc.py" and changed a-b to a+b. Not tested: subagents, resume and limits, or the desktop app. Test environment: Linux x86_64 box with 8 CPUs and no GPU, no vendor account or API key. Agent tools were pointed at a local open-weight model (Qwen2.5-3B-Instruct, 4-bit, served by Ollama 0.40.11), so we tested the tool, not any vendor's hosted models. Full log: /tests/goose. Test scope: - Kind and date: smoke test, 2026-10-25 (America/Chicago). Not a benchmark. - What we ran: Unpacked the official release archive (the install script stopped on a missing bzip2), goose --version printed 1.56.0, and ran 'goose run' with a local Ollama provider; attempt 1 failed to change the file, attempt 2 with exact strings changed a-b to a+b. - Model used: Qwen2.5-3B-Instruct (4-bit), a small local open-weight model served by Ollama 0.40.11; no vendor-hosted model was used - Environment: Linux x86_64 box, 8 CPUs, no GPU; no vendor account, login or API key - Not tested: Subagents, resume and limits, the desktop app. Full test log (commands, versions, raw output): [goose 1.56.0: our test log](https://theagenticweekly.com/tests/goose) Pricing: Not verified. We have not verified pricing for this tool; see the official site. ## Best for - [Best AI tool for AI coding agents](https://theagenticweekly.com/best/ai-coding-agents) (#8): Open-source agent that works with many LLMs and now has resumable subagents. ## What's new goose v1.56.0 focuses on subagents: you can resume failed or cancelled delegated subagents, view subagent sessions in the UI and cap concurrency. It also adds a Windows installer for the desktop app. - v1.56.0 (Oct 23): Resume failed or cancelled delegate subagents. - v1.56.0: Limit concurrent subagents with GOOSE_MAX_CONCURRENT_SUBAGENTS, and open view-only subagent sessions from the UI. - v1.56.0: Automatic thinking level operation for improved model reasoning. - v1.56.0: Windows NSIS installer for the desktop app, and a new IO Intelligence provider. - v1.56.0: Replaces PCTX code mode with a Boa JavaScript runtime. ## How to set it up: Install goose and run your first session Install steps come from the vendor's docs. We ran the install and a short real task on 2026-10-25; see the test notes above for what was and was not covered. 1. **Install the CLI.** The README gives a one-line installer. Read the script first if you prefer. ``` curl -fsSL https://github.com/aaif-goose/goose/releases/download/stable/download_cli.sh | bash ``` 2. **Pick a provider.** Run goose and follow the prompts to connect a model provider. goose works with any LLM according to its description. 3. **Try subagents.** If you use delegated subagents, v1.56.0 lets you cap how many run at once with an environment variable. ``` export GOOSE_MAX_CONCURRENT_SUBAGENTS=3 ``` ## Pros and cons Pros: - Apache-2.0 and very large GitHub following - Desktop and CLI - Subagent controls added in this release Cons: - The project moved organizations, so older links and names may differ - Model quality depends on your provider - With a small local model the first edit failed; we did not test subagents or the desktop app ## Alternatives - [opencode](https://theagenticweekly.com/tools/opencode): Open-source terminal coding agent with a TUI, a desktop app and many model providers. - [Cline](https://theagenticweekly.com/tools/cline): Open-source coding agent as an IDE extension, a CLI and an SDK. - [Claude Code](https://theagenticweekly.com/tools/claude-code): Anthropic's agentic coding tool for your terminal, IDE and GitHub. - [Amazon Q Developer](https://theagenticweekly.com/tools/amazon-q-developer): AWS's AI coding assistant for IDEs and the CLI, with a free tier and a $19 Pro tier; IDE plugin support ends April 30, 2027. - [Augment Code](https://theagenticweekly.com/tools/augment-code): Coding agent and 'Cosmos' platform with a Context Engine, CLI and cloud agents; plans are flat team prices with usage included. - [Devin](https://theagenticweekly.com/tools/devin): Cognition's autonomous software engineer: a cloud coding agent that tests in its own browser and aims to hand back a mergeable pull request. - [Kiro](https://theagenticweekly.com/tools/kiro): AWS's agentic coding tool that turns prompts into requirements, designs and tasks (spec mode), with an IDE and a CLI. - [Qodo](https://theagenticweekly.com/tools/qodo): AI code review platform with 15+ agentic workflows for IDEs, pull requests and security, priced by team credits. - [Warp](https://theagenticweekly.com/tools/warp): Warp's terminal and agent platform; it now pitches 'open infrastructure for cloud software factories' defined as code, with any model. - [Zed](https://theagenticweekly.com/tools/zed): High-performance, multiplayer code editor from the creators of Atom and Tree-sitter, open source with AI optional. - [Codex CLI](https://theagenticweekly.com/tools/codex-cli): OpenAI's open-source coding agent that runs in your terminal. - [Kilo Code](https://theagenticweekly.com/tools/kilo-code): Open-source AI coding agent for VS Code, JetBrains, CLI and Cloud, with 500+ models at zero markup or your own keys. ## FAQ ### What is new in goose v1.56.0? Resumable subagents, a concurrency limit, subagent session links in the UI, automatic thinking level, a Windows desktop installer and a new IO Intelligence provider. ### How do I install goose? The README's CLI install is a one-line shell installer. A desktop app is also available. ### Is goose free? The software is Apache-2.0 open source. Model usage costs depend on the provider you connect. ## What we have not verified - Any pricing from hosted providers - Real-world quality (not tested) - Whether the thinking-level automation helps in practice - Subagent controls, resume and the desktop app (not tested) ## Sources - [goose v1.56.0 release](https://github.com/aaif-goose/goose/releases/tag/v1.56.0) (fetched 2026-10-25) - [goose README](https://github.com/aaif-goose/goose) (fetched 2026-10-25) ## How to cite Agentic Weekly, "goose v1.56.0: what's new, verdict & setup", last updated 2026-10-25, https://theagenticweekly.com/tools/goose Keep the Reported or Tested label when you quote the verdict.