Build the workbench that survives a model switch.
Cloaky is now two products. Base is the free workbench. Every session in it runs on Venice, on the model and privacy tier you pick, with no other provider account involved. Pro is where the multi-provider, local-model and fleet work continues, in a private build. Both keep the project, thread, permissions, review and cost in view while the model changes.
Keep the agent loop useful.
An agent does more than chat. It reads a project, makes a plan, runs tools, asks for permission, changes files and hands work to subagents. Then it leaves a diff someone has to review. Base holds that loop together for everyone, free, and shows the engine, model, permission, file change, and cost at the point they matter. Nothing about that loop is held back for Pro.
The model can change. The work should not disappear.
Base gives everyone one route we can describe honestly. Pro is where the harder routes live: local inference, a provider you already pay for, a fleet of machines. It is being built in private because a month of weekly releases exposed the edges of each. The habits stay the same in both: choose the route, watch the action, review the result, know what the session cost.
The rules that keep the product honest.
The route is part of the result
Every session names the engine, the model and where the inference runs. In Base the answer is always Venice, on the tier of the model you picked.
The agent can stop
Shell commands and other consequential actions can pause with the exact command and scope in view. The important button is sometimes “not yet.”
Review is another turn
The diff belongs in the conversation. Comment on a line, send the request back, or rewind the conversation, files, or both.
Receipts beat vibes
Session cost, tool activity, files changed, tests, open issues, and subagent work should survive “done.”
Privacy claims name their owner
Venice’s tiers, local inference and direct providers are different paths with different owners. We say which one is in use, what its owner promises and where the edge is. A Base page never describes a Pro path as though you had it.
The order of the work matters.
The order we mean to earn things in, with no dates. The changelog records what shipped.
Ship Cloaky Base in the open
Agent sessions with permissions and approval gates, review with line comments, rewind, subagents, bots and squads with routines, and cost readouts. Five engines, all running on Venice models. Free, for macOS and Linux, downloadable here. Base started at v0.1.1 on October 14, 2026; the version number restarted because Base is a new product line.
v0.1.3 · published October 29, 2026Build Cloaky Pro in private
Three months, heads-down, from mid-September. Pro is the full workbench. It has every engine including Cursor and Prime Agent, local models through Ollama and LM Studio, direct providers, and fleet, bots and squads. There is no public download while the build runs. We will describe the features when they exist. A small user study group is being picked from the people who write in. They get Pro, and the request button is at the foot of this page.
Private build · three monthsFind the real privacy boundary
When a product handles code, “private” has to name a mechanism someone can check. We are tracing each path: session turns, headlines, assists, subagents, tools, engine vendors, and account systems. Base has one route to trace and Pro will have several, so the research covers both. The output should be better controls, clearer docs, and fewer claims we cannot prove.
Research underway · evidence before claimsCarry the work forward
Once both products have a shape, the question changes from “which model is best?” to “how does my work survive the next one?” We want the project, decisions, review trail and receipts to stay useful as engines, providers and devices change. They should mean the same thing on Base and on Pro. That is a direction, with no feature promised.
Further out · continuation over noveltyTurn the research into product defaults
Privacy work counts only if it changes the product. That means clearer route choices, safer boundaries, useful warnings, and a better answer when someone asks where a piece of work went. Whatever the research earns should land in Base first, since Base is the version most people will run. There is no final checkbox here. The workbench has to keep earning trust as the intelligence around it changes.
Open-ended by design · product over postureThe roadmap does not end with the split. Every new engine or surface has to improve the same loop. Choose the model, see what it can touch, let it act, inspect what it changed, and keep the record.
Build it, use it, then tell the truth about it.
The public beta ran eleven releases in a month and then closed, because shipping the whole workbench weekly was not making the hard parts right. We ship a surface, use it on real work, measure what it did, write down the limit, and decide what the next build has earned. The labels stay plain: shipped, being tested, under research, or planned.