Freebird AI — AI Coding Assistant for VS Code

No setup. No throttling. 20 free AI edits/day.

AI coding assistant · Copilot alternative · Cursor alternative · multi-file AI edits · AI refactoring · codebase search · Gemini Flash · Kimi K3 · custom LLM providers (OpenRouter, Together, self-hosted) · prompt templates · Ollama · BYOK · local AI · privacy-first

Freebird AI — inline edit rewriting a selection in place

Install Freebird AI and start coding in seconds — no API keys, no throttling, no configuration. You get 20 free advanced AI edits per day powered by Gemini Flash, plus free BYOK and unlimited local AI when you want full privacy.

Copilot throttled? Cursor too expensive? GitHub limits hit? Freebird never blocks you — it picks up where other tools stop.

Upgrade to Pro — $6 USD/month → · or start a free 7-day Pro trial (no card needed) right from the chat panel

⭐ If Freebird saves you time, a GitHub star helps others find it — thank you!

Why Freebird

	Copilot	Cursor	Freebird Free	Freebird Pro
Price	$10/mo	$20/mo	Free	$6/mo
Setup required	No	Yes	No	No
Multi-file agent edits (terminal, checkpoints)	Limited	✅	❌	✅
Local AI (Ollama)	❌	❌	✅	✅
BYOK (Claude, GPT-4o, DeepSeek, Qwen, Kimi K3, custom)	❌	✅	✅	✅
Cloud edits throttled	✅	✅	20/day soft cap	Never

See It in Action

Multi-file agent edit with Approve / Reject (Pro)

Ask Freebird to update your products page, add images to cards, or refactor across files — it shows a full diff and waits for your approval before changing anything.

Freebird AI — multi-file agent edit with approve/reject flow

Agentic page editing across HTML and CSS (Pro)

Freebird reads your existing code structure, understands the context, and makes targeted edits across files in one agent run.

Freebird AI — agentic page editing across HTML and CSS

Free vs Pro

Feature	Free	Pro ($6 USD/mo)
AI chat (unlimited questions)	✅	✅
Active file + `@` file context	✅	✅
`/` slash commands	✅	✅
Works instantly — no setup	✅	✅
Unlimited local Ollama (100% private)	✅	✅
Bring your own API keys — BYOK (Anthropic / OpenAI / DeepSeek / Qwen / Kimi K3 / Custom Provider), direct-to-LLM speed & total privacy	✅	✅
Inline edit (`Ctrl+Alt+K`) & AI commit, cloud-powered	20/day	Unlimited
Multi-file agent edits, terminal commands, checkpoints	—	✅
Project memory across sessions	—	✅
Full (non-lite) Gemini model on cloud edits	—	✅

Pro tip: Connect your DeepSeek API key — it's free on every plan, scores higher than GPT-4o on coding benchmarks, and costs about $0.20/million tokens. Thousands of unthrottled edits a month for pennies, no Pro required.

What Freebird Replaces

GitHub Copilot — when you hit your monthly speed limit
Cursor Composer — multi-file agent edits, without migrating from VS Code
Claude Code — the same bring-your-own-key workflow, free, without leaving VS Code
Local coding agents — Ollama integration built in, unlimited and private
Cloud-forbidden environments — enterprise policies that ban cloud AI, NDA-bound freelance work — Pro's Agent mode runs on Ollama with zero calls ever leaving the machine

Features

Freebird AI feature highlights

Works Immediately — No Setup Required

Install and start coding. Your first 20 advanced edits per day are powered by Gemini Flash — no API key, no Ollama, nothing to configure.

20 Free Advanced Edits Every Day

Unlimited chat, plus 20 cloud-powered inline edits and AI commits a day. Resets daily, no card required. Multi-file agent edits and terminal commands are Pro (see below).

Hit Your Daily Cap? You've Got Options

If you've configured Ollama as your backend, Freebird falls back to it automatically when the cloud quota runs out. Otherwise you'll see a clear "free edits used up" prompt with a no-card 7-day Pro trial one click away — or switch to BYOK (free, unlimited) or Ollama (free, unlimited) anytime from Freebird: Configure AI Backend.

Multi-File Agent Edits with Approve / Reject (Pro)

Freebird reads your codebase, fetches relevant files, and makes targeted edits across multiple paths. Every write shows an Approve / Reject card — nothing changes silently.

Checkpoints — Undo an Entire Agent Turn (Pro)

Every Pro agent turn that creates, edits, copies, or downloads a file gets a "Checkpoint saved" card with a one-click Restore button, reverting every file that turn touched back to its state before the turn. Covers file changes; a turn that also ran a shell command or git push is clearly labeled as not revertible rather than pretending to cover it.

Web Context — Read Any Page the Agent Needs (Pro)

The agent can fetch a webpage's readable content directly — documentation, an article, a URL you paste in — instead of only searching your own codebase. Requests to private/internal addresses (localhost, LAN ranges, cloud metadata endpoints) are blocked, and fetched content is explicitly marked as untrusted reference material before it ever reaches the model, so a malicious page can't pass itself off as an instruction.

Inline Edit — Cursor-style

Select any code, press Ctrl+Alt+K, type an instruction, and the selection is rewritten in place.

Bring Your Own Keys — Unthrottled, free

Plug in your own Anthropic Claude, OpenAI, DeepSeek, Qwen, or Kimi K3 API key — or point Freebird at any OpenAI-compatible Custom Provider (OpenRouter, Together, self-hosted, etc.). Direct-to-LLM speed, total data privacy, no middleman quotas — free for everyone, since the calls never touch Freebird's servers.

Full Agent Mode With Zero Cloud Calls (Pro)

Agent mode isn't locked to Freebird's cloud — it routes through whatever backend you've configured, same as everything else. Set the backend to Ollama and get the full Pro feature set (multi-file edits, terminal commands, checkpoints, project memory) with every request staying on your machine. Built for teams whose policy forbids cloud AI, or freelancers working under an NDA that does the same — the model runs locally, so there's nothing to disclose.

Smart Chat with File Context

Type @filename to inject any file into the conversation. Type / to see all available commands.

Git Integration

Generate commit messages, push to remote, check git status — all from the chat panel.

Project Memory (Pro)

Freebird saves notes about your project to .freebird/memory.md and loads them automatically. Use /memory to see what's saved and /forget to clear it.

Built-in Prompt Templates

Run Freebird: Use Prompt Template to start from a ready-made prompt for a common task — populates the chat input for you to edit before sending, rather than firing immediately. Ships with three: Codebase Cartographer (maps a project's architecture — pairs especially well with a large-context model like Kimi K3), Security Auditor (concrete, cited vulnerability findings — no theoretical padding), and Multi-File Test Engineer (finds untested code paths and writes tests matching your existing conventions).

Pick the Right Model

Model	Best for	Cost
Gemini Flash (built-in)	Default free tier — fast, no setup	Free (20/day)
Ollama (local)	Unlimited local AI — free, 100% private	Free
DeepSeek V4-pro	Advanced reasoning, coding, debugging	~$0.14/M tokens
Qwen 2.5 Coder	High-accuracy coding	~$0.16/M tokens
Kimi K3	Frontier-scale reasoning, huge codebases (1M token context)	See platform.moonshot.ai
GPT-4o	Best all-rounder	~$2.50/M tokens
Claude Sonnet	Complex refactoring & architecture	~$3/M tokens
Custom Provider	Any OpenAI-compatible API — OpenRouter, Together, self-hosted, etc.	Depends on provider

BYOK models are free for everyone (bring your own API key/cost). Gemini Flash and Ollama are also always free. Pro adds Agent mode (multi-file edits, terminal, checkpoints) and unlimited cloud edits on the full Gemini model.

Getting Started

Option 1 — Just Install (Recommended)

Install Freebird AI
Open chat (Ctrl+Alt+O)
Start coding — 20 free AI edits/day, no setup needed

Option 2 — Ollama (Unlimited Free, Local)

Install Ollama
Run ollama pull qwen2.5-coder in a terminal
Run Freebird: Configure AI Backend → select Ollama

Option 3 — Anthropic Claude (BYOK, free)

Get an API key at console.anthropic.com
Run Freebird: Configure AI Backend → select Anthropic Claude

Option 4 — OpenAI (BYOK, free)

Get an API key at platform.openai.com
Run Freebird: Configure AI Backend → select OpenAI

Option 5 — DeepSeek (BYOK, free)

Get an API key at platform.deepseek.com
Run Freebird: Configure AI Backend → select DeepSeek

Option 6 — Qwen 2.5 (BYOK, free)

Get an API key at dashscope.console.aliyun.com
Run Freebird: Configure AI Backend → select Qwen 2.5

Option 7 — Kimi K3 (BYOK, free)

Get an API key at platform.moonshot.ai
Run Freebird: Configure AI Backend → select Kimi K3
Defaults to kimi-k3 (2.8T frontier, 1M token context) — set freebird.model to kimi-k2 for the previous generation instead

Option 8 — Custom Provider (BYOK, free)

Point Freebird at any OpenAI-compatible API — OpenRouter, Together, Groq, Fireworks, a self-hosted vLLM/LM Studio server, or anything else that speaks the /chat/completions format.

Run Freebird: Configure AI Backend → select Custom Provider
Enter the provider's base URL (e.g. https://openrouter.ai/api/v1) and the exact model id it expects (e.g. openai/gpt-4o)
Enter your API key when prompted

Commands

Command	Shortcut	Description
Freebird: Open Chat	`Ctrl+Alt+O`	Open the AI chat panel
Freebird: Edit with AI	`Ctrl+Alt+K`	Inline rewrite selected code
Freebird: AI Commit	—	Generate a commit message
Freebird: Configure AI Backend	—	Switch between Gemini / Ollama / Claude / OpenAI / DeepSeek / Qwen
Freebird: Activate Pro License	—	Enter your Pro license key

Chat Commands

Command	Description
`/commit`	Generate a commit message
`/push`	Push to remote
`/status`	Show git status
`/memory`	Show project memory (Pro)
`/forget`	Clear project memory (Pro)
`/clear`	Clear conversation history
`/help`	Show all commands

How the Agent Works (Pro)

Reads your workspace file tree automatically
Fetches specific files it needs
Searches the codebase for symbols, patterns, or text
Edits files with targeted diffs — Approve / Reject before anything changes
Creates new files — preview shown before creation
Runs terminal commands — shown before execution
Commits and pushes — requires your explicit approval

Nothing is modified silently. You stay in full control.

Settings

Setting	Default	Description
`freebird.backend`	`cloud`	AI backend: `cloud`, `ollama`, `anthropic`, `openai`, `deepseek`, `qwen`
`freebird.apiKey`	(empty)	API key for BYOK backends
`freebird.model`	(auto)	Override the default model
`freebird.ollamaUrl`	`http://localhost:11434`	Ollama server URL
`freebird.licenseKey`	(empty)	Pro license key
`freebird.telemetry.enabled`	`true`	Anonymous usage analytics (no code/PII)

Privacy

Gemini Flash (free tier): messages processed by Google's API. No code stored by Freebird.
Ollama: all processing is local — no data leaves your machine. This applies to Agent mode too (Pro) — set the backend to Ollama and multi-file edits, terminal commands, and checkpoints all run without a single cloud call, which is what makes Freebird usable under a corporate no-cloud-AI policy or an NDA that forbids sending code off-machine.
Anthropic / OpenAI / DeepSeek / Qwen / Kimi K3 / Custom Provider: code sent to that provider's API under your own account.
Freebird AI (Ten Labs Pty. Limited) never collects or stores your code or conversation data.
Web fetches (fetch_url): private/internal addresses (localhost, LAN ranges, cloud metadata endpoints) are blocked, and fetched page content is explicitly marked as untrusted before it reaches the model.
Checkpoint restores can't write outside your workspace folder, even if a checkpoint record were somehow corrupted.

Support

support@ten-labs.com.au — payments, license activation, or anything else.

Contributing

Open source (MIT). Issues and PRs welcome at the GitHub repository.

Freebird AI

Ten Labs