Coding assistants and agents5 projects, checked September 17, 2026

Open-source alternatives to Devin

The open-source alternatives to Devin are OpenHands, goose, Cline, Kilo Code and gptme. OpenHands is the closest in shape: an MIT control centre that runs coding agents on your own infrastructure, in containers, from GitHub issues, schedules and webhooks. goose is a general-purpose agent with a desktop app and an API. Cline runs headless in CI and as scheduled agents; Kilo Code's cloud agent runs from the web; gptme scaffolds a scheduled service on a server. A team switches when the agent must run inside its own network with a model it chooses.

An always-on agent on your own infrastructure
OpenHands
A general agent with a desktop app and an API
goose
Headless runs in CI and scheduled agents
Cline

5 of 5 carry an OSI licence. 5 of 5 run on your own machines. Figures read from GitHub on September 17, 2026.

Pick by need

One project per need, with the reason from its own README and the one thing to know before you choose it.

An always-on agent on your own infrastructure

OpenHands

MIT control centre that runs the OpenHands agent, Claude Code, Codex or any ACP agent in Docker, on VMs or in company infrastructure, with automations from Slack, GitHub and Linear on schedules and webhooks.

Know firstBeta, with a README warning that the npm and source routes give the agent full filesystem access.

OSI88,343 stars

A general agent with a desktop app and an API

goose

Apache-2.0, Linux Foundation governance, 15+ providers including Ollama, 70+ MCP extensions, and an API to embed it.

Know firstNo scheduling or containerised runner described in the README.

OSI54,402 stars

Headless runs in CI and scheduled agents

Cline

Apache-2.0, a CLI that runs with zero interaction and JSON output, scheduled agents with cron, multi-agent teams, and connectors for Slack, Telegram, Discord and Linear.

Know firstRuns where you run it; there is no hosted runner in the open-source project.

OSI68,575 stars

A task run from the web

Kilo Code

MIT, with a hosted Cloud Agent that runs Kilo with no local machine and hosted code reviews.

Know firstThe cloud agent runs on Kilo's infrastructure, not yours.

OSI27,344 stars

A scheduled agent on a server

gptme

MIT, with systemd scaffolding for scheduled autonomous agents, sandboxed execution in Docker or Wasmtime, and a browser, shell and file tools.

Know firstA single-maintainer project with an in-progress list in its README.

OSI4,417 stars

What to take away

  • OpenHands is the closest open-source shape to Devin: assign work, run it in a container, get results in Slack.
  • Every option here is software you run; Devin's infrastructure and hardening are what you take on.
  • None requires a specific model; OpenHands is bring-your-own and goose lists 15+ providers.
  • Read the safety notes in each README before running an autonomous agent with filesystem access.

The projects, side by side

The facts that decide most choices. Each project's page carries the full panel.

ProjectBest forOpennessLicenceRuns asOwn machinesStars
OpenHandsAn always-on agent on your own infrastructureOSIMITnpm package, Docker, From sourceYes88,343
gooseA general agent with a desktop app and an APIOSIApache-2.0Desktop app, CLI, APIYes54,402
ClineHeadless runs in CI and scheduled agentsOSIApache-2.0VS Code extension, JetBrains plugin, CLI, Desktop app, Node.js SDKYes68,575
Kilo CodeA task run from the webOSIMITVS Code extension, JetBrains plugin, CLI, Web (Cloud Agent)Yes27,344
gptmeA scheduled agent on a serverOSIMITCLI (pipx, uv or pip), Web UI, Desktop app, MCP serverYes4,417

Licence and star figures read from GitHub on September 17, 2026. Stars are shown as a dated fact and were not used to rank this page.

Each project, in turn

OpenHands

OSIAn always-on agent on your own infrastructure

An MIT self-hosted control centre, called Agent Canvas, that runs the OpenHands agent or Claude Code, Codex, Gemini and any ACP-compatible agent across local, Docker and cloud backends, with automations triggered by schedules and webhooks.

MIT control centre that runs the OpenHands agent, Claude Code, Codex or any ACP agent in Docker, on VMs or in company infrastructure, with automations from Slack, GitHub and Linear on schedules and webhooks.

Know firstBeta, with a README warning that the npm and source routes give the agent full filesystem access.

Licence
MIT
Runs as
npm package, Docker, From source
Platforms
macOS, Linux, Windows (separate guide)
Hosted option
Sold by the maintainers
Last release
v1.20.0, Sep 17, 2026

goose

OSIA general agent with a desktop app and an API

An Apache-2.0 general-purpose AI agent from the Agentic AI Foundation at the Linux Foundation, written in Rust, with a desktop app, a CLI and an API, 15+ model providers including Ollama, and 70+ extensions over MCP.

Apache-2.0, Linux Foundation governance, 15+ providers including Ollama, 70+ MCP extensions, and an API to embed it.

Know firstNo scheduling or containerised runner described in the README.

Licence
Apache-2.0
Runs as
Desktop app, CLI, API
Platforms
macOS, Linux, Windows
Hosted option
None from the maintainers
Last release
v1.51.0, Sep 17, 2026

Cline

OSIHeadless runs in CI and scheduled agents

An Apache-2.0 coding agent for VS Code, JetBrains, the terminal and a desktop app that edits files as reviewable diffs, runs commands with your approval, works with any model provider including Ollama, and connects tools through MCP.

Apache-2.0, a CLI that runs with zero interaction and JSON output, scheduled agents with cron, multi-agent teams, and connectors for Slack, Telegram, Discord and Linear.

Know firstRuns where you run it; there is no hosted runner in the open-source project.

Licence
Apache-2.0
Runs as
VS Code extension, JetBrains plugin, CLI, Desktop app, Node.js SDK
Platforms
VS Code and JetBrains IDEs, Desktop app for macOS and Windows, CLI anywhere Node.js runs
Hosted option
None from the maintainers
Last release
desktop-v0.0.32, Sep 18, 2026

Kilo Code

OSIA task run from the web

An MIT coding agent for VS Code, JetBrains and the CLI with inline autocomplete, Code, Plan, Ask, Debug and Review modes, an MCP marketplace, access to 500+ models with no API key to start, and a hosted cloud agent.

MIT, with a hosted Cloud Agent that runs Kilo with no local machine and hosted code reviews.

Know firstThe cloud agent runs on Kilo's infrastructure, not yours.

Licence
MIT
Runs as
VS Code extension, JetBrains plugin, CLI, Web (Cloud Agent)
Platforms
VS Code and JetBrains IDEs, CLI binaries for Windows x64, macOS (Apple Silicon and Intel), Linux x64 and ARM
Hosted option
Sold by the maintainers
Last release
v7.7.3, Sep 17, 2026

gptme

OSIA scheduled agent on a server

An MIT personal AI agent for the terminal that runs shell and Python locally, patches files, browses the web, works with Anthropic, OpenAI, Google, xAI, DeepSeek, OpenRouter or a local llama.cpp server, and is both an MCP client and an MCP server.

MIT, with systemd scaffolding for scheduled autonomous agents, sandboxed execution in Docker or Wasmtime, and a browser, shell and file tools.

Know firstA single-maintainer project with an in-progress list in its README.

Licence
MIT
Runs as
CLI (pipx, uv or pip), Web UI, Desktop app, MCP server
Platforms
Anywhere a terminal runs with Python 3.10 or newer, Desktop app for Linux, macOS and Windows
Hosted option
Sold by the maintainers
Last release
v0.33.0, Aug 19, 2026

How to choose

The questions that settle it, in the order they usually come up.

  1. Where should the agent run?

    Your servers in containers: OpenHands. Your laptop: goose or Cline. Someone else's cloud from a browser: Kilo Code's cloud agent. A Linux box on a timer: gptme.

  2. What triggers the work?

    GitHub issues, webhooks and schedules: OpenHands. Cron: Cline or gptme. A person in a chat app: Cline's connectors.

  3. How much autonomy?

    Cline asks on every action by default. OpenHands and gptme offer containers and sandboxes. Kilo Code's --auto removes all prompts.

Where Devin is still the better choice

Devin, in the vendor's own words, is "the AI software engineer" that plans and executes engineering tasks such as code migrations and on-call incident resolution, by Cognition's own description. devin.ai. The cases below are where that still wins.

  • You want an engineer you assign work to and forget about, with the infrastructure, sandboxing and hardening done for you. Every option here is something you run.
  • You want Cognition's own models and its Slack and Linear workflows as a service.
  • You have no one to operate containers and read a hardening guide. OpenHands warns in its README that two of its install routes give the agent full filesystem access.
  • You want one vendor accountable for the agent's results.

How this page was checked

Numbers from GitHub, dated

Stars, licence, last commit and last release are read from the GitHub API by a script every week and carry the date they were read. Nobody types them.

Claims from the README

What each project does is written from its own README and LICENSE, read in full on the date shown, and phrased as the project's claim. The sources are listed at the end.

What is never said

That a project is faster, better or cheaper than Devin. A price for anyone. A user count. A roadmap. If a fact is not on this page, it was not in the source.

OSI Open source (OSI licence)
The code, and the weights where the project is a model, carry a licence on the OSI approved list.
Source available Source available
The code is public, and the licence limits what you may do with it. Read the LICENSE file before commercial use.
Open weights Open weights
The model weights can be downloaded and run. The training data or code is closed, or the weights carry a restriction the code does not.

Questions people ask

What is the best open-source alternative to Devin?

OpenHands (MIT), whose Agent Canvas runs the OpenHands agent, Claude Code, Codex or any ACP agent across local, Docker and cloud backends, with automations from Slack, GitHub and Linear on schedules or webhooks. It is the closest to Devin's always-on shape. goose (Apache-2.0) and gptme (MIT) are general agents that can be scheduled; Cline runs headless in CI.

Can an open-source agent take a GitHub issue and open a pull request?

OpenHands describes automations that decompose GitHub issues into tasks and run in response to webhooks. Gemini CLI, listed under Claude Code alternatives, ships a GitHub Action for pull request review and issue triage. Tabby points to a separate project, Pochi, for turning issues into pull requests.

Which one runs entirely on my own infrastructure?

OpenHands, with its Docker route and its Agent Server REST API for running agents on your own hosts, or on VMs and company infrastructure. gptme scaffolds a systemd service and runs sandboxed in Docker or Wasmtime. goose runs on your machine with local models through Ollama. Kilo Code's cloud agent runs on Kilo's infrastructure.

Is it safe to run an autonomous agent on my laptop?

Read each README. OpenHands says its npm and from-source routes give the agent full access to your filesystem and to be careful, and offers a Docker route confined to a mounted directory. gptme sandboxes execution in Docker or Wasmtime. Cline requires approval on every edit and command unless you enable auto-approve. Kilo Code's --auto flag removes all prompts.

Do these need a specific model?

No. OpenHands is bring-your-own-model. goose works with 15+ providers including Ollama. Cline supports Anthropic, OpenAI, Google, OpenRouter, Bedrock, Azure, Vertex, Groq, Ollama and LM Studio. gptme runs hosted providers or a local llama.cpp server. Kilo Code offers 500+ models through its gateway.

Sources

  1. Devin homepage, the vendor's own description, read September 17, 2026
  2. OpenHands README, read September 17, 2026
  3. goose README, read September 17, 2026
  4. Cline README, read September 17, 2026
  5. Kilo Code README, read September 17, 2026
  6. gptme README, read September 17, 2026

Running one of these inside your own environment, with your own data and your own security rules, is the kind of work Reveneau does. Read how a forward deployed engagement works.