The agents you run.One control plane.Governed, on the record.

Skyflo is the control plane for agentic engineering: a local-first layer above the coding agents, models, and repositories you already use. A persistent mission holds one objective, its approved plan, isolated specialist work on those agents, browser and terminal activity, automations, independent review, an append-only record, and what the work taught.

No. It runs them. Codex, Claude Code, Cursor Agent, Antigravity, Grok CLI, OpenCode, Pi, and conforming ACP agents work inside a Skyflo mission with their existing sign-in, each in an assigned worktree, under one approval policy. Provider-direct Kimi, GLM, Mistral, DeepSeek, and open-weight models use your configured access. Skyflo Managed is planned and is not available yet.

Coding agents execute development tasks. Skyflo is the layer above them that keeps the approved objective, delegated work, approvals, execution evidence, independent review, and learned memory together across agents, repositories, and sessions.

Since Desktop 1.1.0, Skyflo learns from the missions it completes, carries what they teach into future work, turns workflows proven in two repositories into Skills, and can change how it routes work. Skyflo improves how it improves, but approvals, credentials, spending limits, and kill switches sit outside anything it can change, and every change it makes to itself can be reverted.

Yes. You register the checkouts a change touches, and the mission plans across all of them at once. Each coder specialist then works in its own git worktree, so parallel changes cannot collide.

Each mission keeps a record on your Mac of the objective, its plan versions and approvals, every Skyflo tool call and permission decision, the reviewer's findings, and what was learned, with sources. The record is append-only and every event is hash-linked to the one before it, which catches accidental corruption, missing events, and reordering. It is not a defense against someone who can rewrite the whole store.

Skyflo Desktop runs the mission on your linked Mac using your repositories and local credentials. Model calls use a signed-in coding agent or your own provider key.

Every plan, including Free, runs the full control plane on each engineer's linked Mac with their own agents and keys, and every mission keeps its own record. Team administration and policy, organization-wide shared memory, managed inference, and contracted security, deployment, and support controls are Planned and are labeled that way wherever they appear. Book a demo to see the current build run a real objective.

What a completed mission taught, such as a convention you stated or a command sequence the work verified, with a link to its source. Memory the evidence supports is kept automatically and the rest waits for you. You can keep, correct, dismiss, or remove each record, and turn learning off. Organization-wide shared memory is Planned and is labeled that way wherever it appears.