Your agent harnesses in one app
Longwave puts Claude Code, Codex, Cursor, and OpenCode in one app. Choose a harness when you start a thread, and continue your work on the same interface.
One platform for separate harnesses
An agent harness runs the agent loop and connects a model to tools. Longwave supports Claude Code, Codex, Cursor, and OpenCode among other harnesses from the same platform. Plugins, access and interfaces are shared between them so you dont have to constantly switch between different user interfaces.
Use your existing subscriptions
Your $200 Claude Code and Codex subscription already gets you heavily subsidized tokens. Use them on Longwave without giving up any of the subsidization. Longwave is built to give your existing subscriptions more mileage.
Orchestrate different models and harnesses
Each model has their own strength and weaknesses. Longwave is built to support long-running, agent swarms across harnesses and models so you use the best agent for the job. You can, for example, ask GPT-6 Astra on Codex to use Opus 5.5 on Claude Code to generate frontend designs or ask Opus to use Astra as an external reviewer of code changes. You can take this much further where you let the agent themselves decide who they will delegate work to and how they will setup the swarm to accomplish a goal. All of this is visible to you in a simple and intuitive interface.
Opinionated defaults, highly customizable
When you setup Longwave, you get an opinionated platform with built in tools & interfaces for you to get the most value out of your existing subscriptions. You still have the power to customize Longwave by adding your own tools, skills and even building your own plugins within Longwave.
Supported harnesses
Use the harnesses you already have in Longwave. Each thread runs on the harness you choose.
And more coming soon
-
Gemini -
Kimi Code -
Grok Code -
Qwen Code
Questions people ask
What is an agent harness?
An agent harness runs an agent loop and connects a model to tools.
What is harness engineering?
Harness engineering covers the agent loop, tool access, context, file changes, error handling, and permissions around a model.
Is Longwave a Claude Code alternative?
Longwave does not replace Claude Code. It gives Claude Code, Codex, Cursor, and OpenCode one platform.
Can I use Claude Code and Codex on the same platform?
Yes. Longwave supports both in the same app, along with Cursor and OpenCode. Each harness keeps its own tools, access, and rules.
What is harness in AI?
A harness is the runtime around a model. Agent harness is the same idea.
What is the best agent harness?
There isn't a single best. Claude Code, Codex, Cursor, and OpenCode differ in tools and limits. Longwave runs all four.
Is Claude Code an agent harness?
Yes. It is Anthropic's coding agent harness.
What's the difference between CLAUDE.md and AGENTS.md?
Claude Code reads CLAUDE.md. Codex and OpenCode read AGENTS.md. Cursor uses its own rules files, and AGENTS.md when that file is there.
What is an AGENTS.md file?
AGENTS.md is a standing instructions file in the repo. Codex and OpenCode read it. Cursor reads it when the file is there. Claude Code reads CLAUDE.md instead.
Can I use a Claude Max plan and a ChatGPT Pro / Codex plan in the same app?
Yes. Longwave signs into your Claude Max and ChatGPT Pro subscriptions and runs Claude Code and Codex on those plans. The included tokens stay with those subscriptions.
Can a thread on one harness delegate work to a different harness?
Yes. A thread on Claude Code, Codex, Cursor, or OpenCode can send a task to a different one. You name the harness, model, and account. The parent decides what context the child gets. That can scale into a swarm across harnesses and models.
Can I run an agent swarm across different harnesses?
Yes. The parent starts children on Claude Code, Codex, Cursor, or OpenCode and chooses what context each one gets. That is a swarm across harnesses and models.
Does Codex or Cursor read CLAUDE.md, or do I need AGENTS.md too?
Codex and OpenCode do not read CLAUDE.md. Longwave does not turn one file into the other, so keep AGENTS.md in the repo if those harnesses work there.
Do I have to pick Claude Code, Codex, or Cursor?
No. Each thread runs on the harness you choose. Claude Code, Codex, Cursor, and OpenCode can run in the same app on different threads.
What do I do when Claude Code hits a usage limit?
Start a new thread on a harness that still has quota, or set it up to fail over automatically. Longwave shows the limits. A started thread stays on the harness it started on.
Can I use Claude Code inside Cursor?
Longwave does not install Claude Code into Cursor. Cursor is one harness and Claude Code is another. You pick which one a thread runs.
If two harnesses work the same repo, do they collide?
Longwave uses workstreams to make sure work done by two harnesses does not collide. Workstreams are git worktrees and branches coupled together, so you do not have to think about which branch is checked out where.
Use the harness you already pay for.
Claude Code, Codex, Cursor, and OpenCode in one Longwave platform.