Clixad
Freemium
Developers who prefer terminal workflows
OpenCode
FreeFree (MIT). Optional Zen models pay as you go; Go $10/mo, Go Plus $40/moOpen SourceAPISelf-HostedAgentic
An open-source terminal coding agent you point at any model, from a hosted API to a model on your own machine.
Similar tools
ClixadFreemiumClixad is a terminal coding agent with free daily usage, configurable approval modes and…
AiruncodeFreemiumAiruncode runs agents locally, lets you plug in your own provider keys or on-device models…
100xpromptFreemiumThe vendor positions this as sovereign AI infrastructure — meaning the compute, the model, and…
Command CodeFreemiumThe agent runs in three modes — interactive CLI, headless with a prompt flag for scripted…
Get This Tool
License: MIT
Local-run terms: Install and run via provided curl, npm, bun, brew, or paru commands; desktop beta available for download; open source with no vendor-hosted requirement.
Pricing
- Model
- Free
- Price
- Free (MIT). Optional Zen models pay as you go; Go $10/mo, Go Plus $40/mo
Summary
OpenCode is an MIT-licensed coding agent for the terminal, with a desktop app (in beta) and an IDE extension. You choose the model provider, including a local model behind an OpenAI-compatible server such as Ollama. We installed version 1.19.2 with its one-line script, which took 3 seconds, and gave it the same three jobs as Goose and Hermes Agent. With Claude Opus 5.5 it passed all 9 runs for $0.19, between Goose ($0.14) and Hermes Agent ($0.27), and it was the fastest of the three on the CSV task. With a 3B local model it passed 3 of 9: the create-file task every time, but none of the report summaries or CSV counts. On every run it tried to reach models.opencode.ai and registry.npmjs.org. It kept working when we blocked them, but it downloads provider packages on first use, so pre-install them if you run it offline.
Bottom line: Pick it if you want an open-source terminal agent and plan to use a capable model; a 3B local model was not enough in our test. Offline, pre-install its provider packages.
We installed and ran it
- How we installed it
- Vendor one-line install script
- Install script run time
- 3.1 s
- Docker image size on disk
- 0.57 GB
- Local model we used
- qwen2.5:3b via Ollama
- Agent tasks passed with the local model
- 3 of 9
- Outside hosts it tried to reach (we blocked them)
- models.opencode.ai, registry.npmjs.org
| Task (qwen2.5:3b) | Passed | Median time |
|---|---|---|
| Create a file with given text | 3 / 3 | 20.8 s |
| Read a report and write a 2-sentence summary | 0 / 3 | 37.7 s |
| Count the data rows in a CSV with a shell command | 0 / 3 | 26.2 s |
With Claude Opus 5.5 (Anthropic API)
- Model
- Claude Opus 5.5 (claude-opus-5-5)
- Agent tasks passed
- 9 of 9
- API cost for all 9 runs (list price)
- $0.19
- Tokens per task (input + output)
- ~29,200
- Tokens per model call
- ~10,500
- Model calls, all 9 runs
- 25
- Input tokens read from the prompt cache
- 95%
- Other hosts it tried to reach (the proxy refused them)
- models.opencode.ai, registry.npmjs.org
| Task (Claude Opus 5.5) | Passed | Median time |
|---|---|---|
| Create a file with given text | 3 / 3 | 11 s |
| Read a report and write a 2-sentence summary | 3 / 3 | 16.3 s |
| Count the data rows in a CSV with a shell command | 3 / 3 | 10.9 s |
Same three tasks, Claude Opus 5.5 driving each agent
| Measured | OpenCode | Goose | Hermes Agent |
|---|---|---|---|
| Tasks passed with Claude Opus 5.5 (of 9) | 9 | 9 | 9 |
| Median time: Create a file with given text (s) | 11 | 8.1 | 18.8 |
| Median time: Read a report and write a 2-sentence summary (s) | 16.3 | 10.3 | 24.8 |
| Median time: Count the data rows in a CSV with a shell command (s) | 10.9 | 11.6 | 22.3 |
| Model calls, all 9 runs | 25 | 26 | 24 |
| Tokens per task, input + output (approx.) | 29,200 | 7,800 | 47,100 |
| API cost, all 9 runs (US$, list price) | 0.19 | 0.14 | 0.27 |
| Outside hosts it tried to reach besides the API (blocked) | 2 | 1 | 0 |
| For reference: tasks passed with qwen2.5:3b on 3 CPU cores (of 9) | 3 | 5 | 0 |
All three ran on October 16, 2026, one after another, with Claude Opus 5.5 (claude-opus-5-5) through Anthropic's API and each agent's default settings: three runs per task, each in a fresh container, graded on the files left behind. The containers had no internet; a proxy let them reach api.anthropic.com and nothing else. Times include the round trips to Anthropic's API, so they show the agent and the API together, not this machine's speed. Tokens come from each agent's own output (Hermes Agent's model-call count is approximate); cost is those tokens at Anthropic's list price. The qwen2.5:3b row is from the earlier local-model runs (Goose and Hermes Agent on October 15, OpenCode on October 16).
What we noticed
- Its qwen2.5:3b run was on October 16, a day after Goose's and Hermes Agent's, so read that row of the comparison table as a reference, not a same-day result. All three Claude Opus 5.5 runs were on the same day. With Claude Opus 5.5, OpenCode passed all 9 tasks.
- It tried to reach models.opencode.ai and registry.npmjs.org with both models (102 lookups over 9 runs with the local model). We blocked both and OpenCode kept working. For the Claude run we pre-fetched its Anthropic provider package into the image, which OpenCode otherwise downloads on first use.
- In the 3B test Ollama capped the context at 32,768 tokens (we asked for 65,536).
Test machine: 8-vCPU AMD EPYC 9354P cloud VM, each test limited to 3 cores and 10 GB RAM, no GPU, no internet (isolated Docker network). Times in seconds are from this CPU-only machine; a desktop with a GPU will be much faster. The VM is shared, so text-generation speed is shown only as a ratio between tools run back to back, not as tokens per second. Docker image size is the on-disk size from docker image ls for every tool. How we test · raw data
Find your fit
Before you choose OpenCode
More options in CLI Coding Agents. Compare pricing, platforms, and features side by side.
Airuncode
Freemium
Developers preferring fully local agent execution
Pros
- Passed all 9 of our agent tasks with Claude Opus 5.5.
- Works with the provider you choose, including a local model behind an OpenAI-compatible server.
- Open source under the MIT license.
Cons
- With a 3B local model it passed 3 of 9 tasks (only the simplest one), so plan on a larger model.
- It reaches out to models.opencode.ai and registry.npmjs.org, and downloads provider packages on first use, so offline setups need those pre-installed.
- The vendor lists its desktop app as beta.
About
- Platforms
- Terminal, Desktop (beta macOS/Windows/Linux), IDE extension
- API Available
- Yes
- Self-Hosted
- Yes
- Last Updated
- June 11, 2026
Best For
Who it's for
- Developers who want an open-source coding agent in the terminal
- Teams that want to choose their own model provider
What it does well
- Write and edit code from the terminal or your IDE
- Run several agents on one project at once
- Share a session with a link
- Sign in with an existing GitHub Copilot or ChatGPT plan
Integrations
Frequently Asked Questions
- Is OpenCode free?
- OpenCode offers a free option. Hosting, model usage, or optional services may have separate costs; check the pricing details and official terms.
- Is OpenCode open source?
- Yes. OpenCode is open source — the source repository is at https://github.com/sst/opencode.
- Does OpenCode have an API?
- Yes. OpenCode exposes a developer API. See the official documentation at https://opencode.ai for details.
- Can I self-host OpenCode?
- Yes. OpenCode supports self-hosting on your own infrastructure.
- What platforms does OpenCode support?
- OpenCode is available on: Terminal, Desktop (beta macOS/Windows/Linux), IDE extension.
People also compared
Curated lists that include this category
OpenCode is an open-source AI coding agent that works in your terminal. It is MIT-licensed. You connect it to a model provider of your choice, including a local model through an OpenAI-compatible server such as Ollama, and it reads, edits and runs code in your project. The vendor also offers a desktop app and an IDE extension.
We installed version 1.19.2 with the vendor’s one-line install script in an isolated container and gave it the same three tasks as goose and Hermes Agent: create a file, summarize a report, and count rows in a CSV with a shell command. With Claude Opus 5.5 through Anthropic’s API it passed all 9 runs. With a small local model (qwen2.5:3b on 3 CPU cores) it passed 3 of 9: the create-file task every time, but none of the report summaries or CSV counts, where the small model’s answers were wrong. Goose and Hermes Agent also struggled with that model, so read it as a limit of the 3B model.
On every run it tried to reach models.opencode.ai and registry.npmjs.org. We blocked both and it kept working. For the Claude run we added its Anthropic provider package to the image in advance, because OpenCode otherwise downloads it the first time you use that provider. Expect that if you run it on a network without internet access.
Contribute
- Work on OpenCode? Claim this listing.
- Embed a badge for your README or site.
Sign in to rate this tool or suggest an edit.