Choose Picasso.
You want the whole agent in one place: memory, tools, subagents, providers, goals, and trust controls.
Claude-style depth. More memory. More choice. No Picasso subscription.
Claude-style depth. More memory. More choice. No Picasso subscription.
picasso auth provider add anthropic --subscriptionpicasso memory import CLAUDE.mdpicasso plan "run the same task Claude Code handled last week"Claude Code is excellent when the answer is Claude. Picasso is for the teams that want the same seriousness with memory that lasts, providers that move, and a price that starts at zero.
You want the whole agent in one place: memory, tools, subagents, providers, goals, and trust controls.
You want its native surface and do not need Picasso's broader harness.
Start with one repository. Keep your current tool. Let the work decide.
| Capability | Claude Code | Picasso |
|---|---|---|
| Context window. | Anthropic only. | Up to 1M context across frontier models. |
| Plan mode. | Yes. | Yes. |
| Auto-diagnostics on edits. | Partial. | Attached to create_file and str_replace across TS, JS, Python, JSON, and Swift. |
| Git worktree isolation. | Flag or manual. | /worktree create|apply|remove and Mac Worktree mode. |
| Local branch review. | Prompted. | /diff main and /review-branch. |
| MCP. | Yes. | Native client and server. |
| Persistent memory. | Project files and session context. | Vector plus graph retrieval across sessions. |
| Goal mode. | Anthropic only. | Every supported provider. |
| Provider routing. | Claude family. | 1-step /model and /provider across Anthropic, OpenAI, Google, xAI, Microsoft Foundry, Qwen, Xiaomi MiMo, Kimi, DeepSeek, Zhipu GLM, MiniMax, OpenRouter, Ollama, vLLM. |
| Tool price. | Included with Claude plans ($20-200 per month). | $0 — bring a plan, a key, local models, or managed access. |
Picasso puts the loop at the center. Plan. Act. Observe. Reflect. Remember. The important state is visible, inspectable, and carried forward.
A plan shows the files, risks, checks, and open questions.
File edits with auto-diagnostics, git worktree isolation, MCP servers, and subagents stay inside the same loop.
Memory keeps the facts and decisions worth bringing back.
Switch models and providers in one step with /model and /provider across 1M-context frontier models and local runtimes.
Run parallel threads in isolated git worktrees via /worktree or Mac Worktree mode, with automatic syntax and typecheck feedback on every file edit.
See what the agent knows. Edit it. Delete it. Bring it back next session.
A long-running goal should not belong to one vendor.
If your organization has standardized on Claude and Anthropic policy, Claude Code keeps the stack simple.
Organizations with a strict single-vendor model policy may want the vendor's own CLI.
picasso auth provider add anthropic --subscriptionpicasso memory import CLAUDE.mdpicasso plan "run the same task Claude Code handled last week"picasso goal create "finish the migration with tests green"Compare model lock-in, memory depth, local execution, pricing path, and trust boundaries in one table.
Run Picasso against a task your current agent already handled and compare the plan, diff, checks, and replay.
Read exactly what Picasso sees and what it never touches.
Picasso for Mac is almost here — a coding agent that looks the way serious tools should, and costs what creative freedom should: nothing. Leave your email and be first on the canvas.
Sponsors and labs — the early canvas is yours. Choose Sponsor or Lab above and we'll reach out before launch.