Ling 3.1 Flash
ling-3.1-flash
Fast everyday model with a 262K-token context. The Harness default.
Unlimited
- IN / 1M
- Free
- OUT / 1M
- Free
- LIMIT
- None
Free desktop coding agent, no card needed
Unlimited AI for $0.
Plus $50 a week for frontier models.
Ling 3.1 Flash, Muse Spark 1.3 and GLM 5.3 Flash have no cap. Every Monday you also get $50 to spend on Opus 5.5, GPT 6.1 Sol, Kimi K3 and 7 more, all inside the FreeFlow Harness.
- Unlimited
- 3
- models, no cap
- Models
- 13
- in one app
- Mondays
- $50
- frontier credit
- Reads your code
- Edits with diffs
- Runs your tests
- Asks before acting
One Harness, every model
- Ling 3.1 FlashinclusionAI
- Muse Spark 1.3Meta
- GLM 5.3 FlashZ.ai
- Opus 5.5Anthropic
- Sonnet 5.5Anthropic
- Haiku 5.5Anthropic
- GPT 6 AstraOpenAI
- GPT 6.1 SolOpenAI
- GPT 6 LunaOpenAI
- Kimi K3Moonshot AI
- DeepSeek V4.1 FlashDeepSeek
- Qwen 3.8 MaxAlibaba Qwen
- Mistral Large 4Mistral AI
UnlimitedPaid from your $50 weekly credit
From sign-up to your first fix in three steps
No setup beyond an install. The Harness handles models, keys and credit.
Create a free account
An email and a password. No card, no trial clock. You get your first $50 straight away.
Install the Harness
Download it for Windows, macOS or Linux, paste the API key from your dashboard, and open a project folder.
Ask for a change
Describe what you want. The agent reads your code, proposes edits and runs your tests, asking before anything changes.
Models to start with
Unlimited models for everyday work, frontier models when it matters. Same Harness either way.
Default
Muse Spark 1.3
muse-spark-1.3
Meta's Muse model, tuned for coding and agent work.
Unlimited
- IN / 1M
- Free
- OUT / 1M
- Free
- LIMIT
- None
GLM 5.3 Flash
glm-5.3-flash
Quick GLM model for small edits and questions.
Unlimited
- IN / 1M
- Free
- OUT / 1M
- Free
- LIMIT
- None
Opus 5.5
opus-5.5
Anthropic's most capable Opus model, for hard and long tasks.
$50/week credit
- IN / 1M
- $4
- OUT / 1M
- $20
- $50 BUYS
- 2.5M out
GPT 6.1 Sol
gpt-6.1-sol
OpenAI's everyday reasoning tier.
$50/week credit
- IN / 1M
- $2
- OUT / 1M
- $10
- $50 BUYS
- 5M out
Kimi K3
kimi-k3
Moonshot AI's latest Kimi model.
$50/week credit
- IN / 1M
- $3
- OUT / 1M
- $15
- $50 BUYS
- 3.3M out
A coding agent that asks before it acts
The Harness works in the project folder you open. It reads and searches on its own, and stops to show you every edit and command before it happens.
- Sees the whole project. Lists, reads and searches files to understand your code before changing it.
- Edits with a diff. Every change shows what's added and removed. Allow it, deny it, or allow the rest of the task.
- Runs your tools. Tests, builds and linters run in your folder, with the output right in the chat.
Permissions
What the agent can do without asking
- Read filesAutomatic
- Search the projectAutomatic
- Edit filesAsks first
- Run commandsAsks first
Edit src/date.ts
- return `${y}-${m}-${d}`
+ return `${y}-${pad(m)}-${pad(d)}`
DenyAllow
Everything you expect from a coding agent
The tools the paid agents have, without the subscription.
13 models, one picker
Opus 5.5, GPT 6.1 Sol, Kimi K3, DeepSeek, Qwen and more. Switch mid-task when a job needs more muscle.
Asks before it acts
Every edit shows a diff and every command waits for your OK. Or let it run on its own in Full auto.
Undo a whole task
Checkpoints save your files before the agent touches them. One click puts everything back.
Knows your project rules
Reads AGENTS.md (or CLAUDE.md) from your project, so it follows your conventions from the first message.
@-mention files
Type @ to hand the agent exactly the files that matter, instead of hoping it finds them.
Plan mode
Have it read the code and propose a plan without changing a thing. Approve, then let it build.
See what every task costs
A live token and credit meter on each task, and a full usage history in your dashboard.
Runs your tools
Tests, builds and linters run in your project folder, with output streamed into the chat.
Any editor, any language
The Harness works on your project folder, next to VS Code, JetBrains, Vim or whatever you use.
The same frontier models, without the monthly bill
What it costs to get started with each tool, checked October 2026.
| Tool | Free tier | Paid plans | Details |
|---|---|---|---|
| FreeFlow | 3 unlimited models + $50 of frontier credit every week | None. It's free | |
| Cursor | Hobby plan with limited agent requests | Pro at $20/month | vs Cursor |
| Claude Code | Not included in the free Claude plan | Claude Pro at $20/month | vs Claude Code |
| Cline | The extension is free; you pay for the model usage | Pay your model provider per token | vs Cline |
| GitHub Copilot | Copilot Free with limited features and 2,000 completions a month | Pro at $10/month | vs GitHub Copilot |
How is it free?
The Harness shows a small sponsor card. Sponsors pay for the tokens, so you don't. That's also why model requests run in the Harness instead of a bare API: the ads are what keep every model free.
Ads are chosen without your personal data, and they never appear inside your code or answers.
Sponsored
Advertise on FreeFlow
Reach developers while they build. Your ad here keeps FreeFlow free.
How the sponsor card looks in the Harness sidebar
Questions
What happens when my $50 runs out?
The weekly-credit models pause until Monday at 00:00 UTC. Ling 3.1 Flash, Muse Spark 1.3 and GLM 5.3 Flash keep working the whole time.
Does unused credit carry over?
No. Every Monday your credit goes back to $50, whether you used $2 or all of it.
How much does a request cost?
Each request costs its tokens at the rates on the models page. A short task on Opus 5.5 costs a few cents; reading a large codebase costs more. Your dashboard shows every request.
Is my code stored?
FreeFlow doesn't keep your prompts or code. They pass through to the model you chose and back. We record token counts so we can track your credit.
Can I call the API from my own code?
Your key can list models and check your balance from anywhere. Model requests run only in the Harness, because the ads there pay for them.
Which systems does the Harness run on?
Windows, macOS and Linux. It works on any project folder you open, in any language.
Is FreeFlow a free alternative to Cursor or Claude Code?
For agent work, yes. FreeFlow reads your project, edits files and runs commands like those tools do, and it's free: three unlimited models plus $50 of frontier-model credit every week. It runs as its own app next to your editor rather than replacing it.
Why are there ads?
A small sponsor card in the Harness sidebar pays for everyone's tokens. That's the whole business model: no subscriptions, no selling your data, and no ads inside your code or answers.
Your first $50 is waiting.
Free account, no card. Takes under a minute.