ANTECANTMake the purpose explicit.
Carry the outcome, constraints, and decisions into the next step.
The control layer for AI work
In developmentOrder before
autonomy.
Put purpose, permission, and proof around the work you delegate to AI.
Antecant is building a local-first control layer for people directing AI work.
Built for the person
responsible for the result.
Local-first by design.
Independent of any one model.
01 / The missing layer
A capable agent still
needs a clear brief.
The goal changes. A session ends. A second agent starts. Somewhere along the way, the decision and the work drift apart.
Antecant is being built to hold the operating record around that work: what matters, what was decided, what may happen next, and what supports the result.
For founders and technical operators, that means a durable place to direct work across changing tools and sessions.
Why this layer mattersSet the limits of action.
Define scope and approval before consequential work begins.
Ask what proves the result.
Keep the evidence beside the claim so it can be checked.
02 / From intent to evidence
Give the work
something to answer to.
Explore a simple example of the intended workflow. This is an interactive concept, not a live agent run.
ILLUSTRATIVE WORK BRIEFPURPOSE
Prepare a product release.
Make the release reviewable before it reaches customers.
- Outcome
- A tested change and a release proposal.
- Constraints
- Keep existing customer data intact.
- Acceptance
- Tests, a readable diff, and remaining risks.
A useful brief preserves what a new session needs to know.
Design direction · Execution controls and evidence workflows are planned. See what exists today ↗
03 / A record that stays
Change the model.
Keep the thread.
The tools will change. Your purpose, decisions, and history should remain yours.
The case for continuityModelsHarnessesTools
ANTECANTThe durable operating recordPurposeDecisionsAuthorityEvidence
Execution is replaceable. The record is the foundation.
04 / Built in deliberate steps
A clear direction.
An honest starting point.
The foundation is taking shape.
Local threads, working briefs, and context records are implemented. Governed execution and evidence-based review are the next layers being developed.
Antecant is not yet a generally available product. We’re interested in concrete workflows from people who need this kind of control.
Read the development statusA few useful distinctions
Before you
begin.
What is Antecant?
A local-first control layer in development for people directing AI work. Its purpose is to preserve the brief, define the authority around execution, and keep the evidence needed to assess an outcome.
Does it replace my coding agent?
The product is designed to sit above replaceable models, tools, and harnesses. Provider integrations and governed dispatch are still planned; this site does not promise compatibility with every agent today.
Can I use it today?
Antecant is in early development. There is no public production download yet. See the current foundation or tell us about your workflow.
Does local-first mean nothing leaves my machine?
No. The canonical work record is designed to stay local, but using an external model or tool can send permitted data to that service. Those transfers need explicit boundaries. Local storage alone does not make remote execution private.
How do you pronounce it?
AN-teh-kant. Antecant is a coined name for the conditions established before autonomous action. The short version: order before autonomy.
For the work that matters
Govern what acts
on your behalf.
Bring a real workflow.
Help shape the layer around it.