Free coding agent · free AI models · sponsor-funded

Code Beautifully.

The most beautiful coding editor ever made — with a serious agent inside, and the AI runs on us. No API key, no subscription, no per-token bill: sponsors cover the inference so you never do. Read-only by default, every write gated behind your approval. That's what Code Freely means.

Not live yet. Free for developers when it is — that's the point.

$brew install picasso waitlistfree editor · free AI · $0 tokens
picasso-core/Sources/ App.swift
Read-only
P
Picasso
New run
Search
Approvals1
Skills
Projects
picasso-core
picasso-ui
harmonic-keyboard
App.swift
GuardChain.swift
This file is open read-only. Picasso edits are gated — request write access to make changes.
12345678910111213141516
import SwiftUI
 
// Entry point for the Picasso desktop shell.
@main
struct PicassoApp: App {
@StateObject private var store = RunStore()
 
var body: some Scene {
WindowGroup {
StudioView()
.environmentObject(store)
.frame(minWidth: 880, minHeight: 560)
}
.windowStyle(.hiddenTitleBar)
}
}
SwiftUTF-8Ln 5, Col 1read-onlyspaces: 2
Assistant
Ask about App.swift — explain, refactor, or generate a patch. Picasso stays read-only until you approve a write.
Picasso
This file wires the SwiftUI App entry point. Lines 5–9 build the window group; the run store is injected as an environment object.
Explain selection Suggest a refactor Find a bug
Ask Picasso about this file…
Approval queue 1 pending
- let hash = bcrypt.hash(pw)
+ let hash = argon2id.hash(pw)
RejectRevise✓ Approve
$0
in AI tokens, ever. Sponsors cover the inference, not you.
no key
no subscription, no credit card. One command and you're coding.
100%
of writes route through approvals by default. Loosening that is your call.
/goal
autonomous work with criteria, budget, approvals, and replay evidence.
How Picasso is free

The AI runs on us.

Every other agent bills you for the model — a subscription, your own API keys, or a meter ticking on every token. Picasso doesn't. Sponsors fund the inference, so you get frontier models inside a beautiful editor for exactly zero.

Bring nothing.

No API key, no subscription, no card. Install in one command and the agent is already wired to real models.

Sponsors cover the tokens.

Inference is paid for by clearly disclosed sponsors — honest, labeled placements, never your wallet. That's what funds the free tier.

Walled off from your work.

Sponsors never see your code and never influence a single token of model output. Your code is never training data or sponsor inventory.

Prefer your own? Drop in an API key or run local models anytime — the free path is a choice, not a cage.

Inference this month Sponsor-funded
ModelFrontier · 200K ctx
Tokens in4.21M
Tokens out1.08M
Agent requests1,284
Your billstandard rate · $63.40
$0.00
covered by sponsor
THE TRUST LOOP

Save routes through
approvals. Always.

01 · You save
The write is queued in the editor — held, not applied. Nothing has touched disk.
02 · It enters the queue
The exact diff lands in the approval queue. Reject, revise, or approve — your call.
03 · Written to disk
Only on approval is the change written — and recorded, with replay evidence.
picasso-core/Sources/ App.swift
Read-only
App.swift
GuardChain.swift
This file is open read-only. Picasso edits are gated — request write access to make changes.
12345678910111213141516
import SwiftUI
 
// Entry point for the Picasso desktop shell.
@main
struct PicassoApp: App {
@StateObject private var store = RunStore()
 
var body: some Scene {
WindowGroup {
StudioView()
.environmentObject(store)
.frame(minWidth: 880, minHeight: 560)
}
.windowStyle(.hiddenTitleBar)
}
}
SwiftUTF-8Ln 5, Col 1read-onlyspaces: 2
Assistant
Ask about App.swift — explain, refactor, or generate a patch. Picasso stays read-only until you approve a write.
Picasso
This file wires the SwiftUI App entry point. Lines 5–9 build the window group; the run store is injected as an environment object.
Explain selection Suggest a refactor Find a bug
Ask Picasso about this file…
Approval queue 1 pending
- let hash = bcrypt.hash(pw)
+ let hash = argon2id.hash(pw)
RejectRevise✓ Approve
Memory

The agent does not
start cold.

Picasso retrieves relevant memory before planning, injects it into context before it acts, and writes back what mattered after the run — facts, style, decisions, failures.

Runtime wired. CLI, TUI, and Mac runs retrieve and write memory.
Retrieval only. Your code is never training data or sponsor inventory.
User-visible. Inspect, edit, pin, export, delete, or clear records.
Failure aware. Repeated mistakes become memories the next run avoids.
What I remember about this codebasefilter: argon2
Facts34 entries · workspace-bound
auth module owns session tokens, JWT verification, 2FA.
Drizzle ORM over Postgres. Migrations in migrations/.
Decisions9 entries · workspace-bound
Chose argon2id over bcrypt. Memory-hardness required by audit.
Bridge bcrypt verifies for 90 days, then auto-migrate on next login.
Style18 entries · account-wide
async/await preferred. Reject .then() chains.
/goal

Give long work
a finish line.

Hand Picasso an objective, completion criteria, permissions, and a budget. It plans, acts, observes, reflects, repairs failures, and keeps going until the work is done or a boundary is reached.

Deterministic criteria. It checks its own completion contract before calling anything done.
Independent audit. A separate pass verifies the result against the goal.
Continues anywhere. Start a goal in the terminal, watch it finish on the Mac app.
migrate-to-argon2idrunning · 3h 22m
Migrate auth module to argon2id, tests green.
COST
$1.84 / 5.00
STEPS
287 / 1000
DONE
62% 2 / 3
bash · pnpm test src/auth/session.test.ts
verify · accepts current session 8ms
hash · argon2id round-trip 112ms
· running migration compat suite
The harness

Plan. Act. Observe. Reflect.

The agent loop is visible by design. You see the plan, the tools, the failures, the checks, and the reason the work changes direction.

01

Read-only plan mode

Architecture, entry points, data flow, risks, and verification — grounded in your real workspace, with writes blocked until you approve execution.

02

Subagents & replay

Named and forked workers investigate in parallel, then replay shows the plans, tools, approvals, edits, and results.

03

Modes, MCP & hooks

Permission mode, MCP tools, hooks, sponsor boundaries, and provider routing stay visible instead of hiding inside a prompt.

Compare

The important state
stays visible.

Plans, goals, memory, subagents, permissions, budgets, and replay are inspectable in the terminal, the TUI, and the Mac app.

Claude CodeCodex CLICursorDevinPicasso
Read-only plan modeYesYesYesYesYes
SubagentsYesLimitedYesYesNamed + fork
Runtime memoryProject filesNoPartialWikiRetrieve · inject · write back
Autonomous /goalAnthropicOpenAINoYesAll routes
Replay evidencePartialPartialNoHostedCLI · TUI · Mac
Free pathTrialLimitedTrialCreditsSponsor-supported
Tool subscriptionClaude plansChatGPT plans$20/moFrom $20/mo$0
Install

Four ways in. Pick yours.

Managed inference, a supported provider subscription, your own API keys, or local models through Ollama and vLLM. macOS, Linux, and Windows.

One binary.
No subscription required.

Zero account to start — bring your own keys, or unlock managed inference through honestly disclosed sponsors.

macOS 13+ · Linux x64/arm64 · Windows 11 · ~12MB · no telemetry by default
brewbrew install picasso
npmnpm i -g @picasso/cli
pnpmpnpm add -g @picasso/cli
windowswinget install picasso.picasso

Install opens the day the doors do.

Code Freely.
Code Beautifully.

Picasso for Mac is almost here — a coding agent that looks the way serious tools should, and costs what creative freedom should: nothing. Be first on the canvas.

Not live yet. Free for developers when it is — that's the point.