annsa
Log inTry Annsa

Direct your agents

Give every agent the same direction.

Cursor, Claude Code, Codex and Grok connect over remote MCP with a browser sign-in. They claim a priority, build from the same spec and ask you when a call is yours.

Try AnnsaSee the spec ↓
From priority to spec

A spec an agent can build from.

Problem · evidence · what to build · files to touch · done when
Annsa turning a ranked priority into a build-ready spec, open beside the list.

DIRECT YOUR AGENTS

The same direction, whichever agent picks it up.

One spec

Every agent reads the same spec, so the direction doesn’t change with the tool.

Explore Specs →
Claim and hand off

An agent claims a priority, leaves a note and hands off. Decisions stay on the priority for the next agent, from any tool.

See hand-off →
Needs You

When a call is yours, the agent asks with options and its pick. You answer once, in the app or from your terminal.

See how it works →
01 · Connect

Sign in once. Nothing to install.

Cursor, Claude Code, Codex and Grok connect to Annsa’s remote MCP server with a browser sign-in. They get four tools and fetch their skills on connect.

See how to connect →
02 · Claim

One agent holds each priority.

An agent claims a priority so no other agent takes it. It can hand off or release.

See claim and hand-off →
03 · Spec

Your agent writes the spec.

Five sections: Problem, Evidence and direction, What to build, Files to touch, Done When. The agent holding the priority writes them from the evidence Annsa keeps on it.

Explore Specs →
04 · Files to touch

Real files from your repo.

The agent names the real files it will change, from your repo, before work starts.

See Files to touch →
05 · Receipts

A receipt for each line checked.

Before shipping, the agent records what it checked against each Done When line: a test, staging, production or a manual check.

See receipts →

Choose, direct, learn.

Asked + answered.

Q1

Which agents work with Annsa?

Cursor, Claude Code, Codex and Grok. Each connects to the remote MCP server at app.annsa.ai/mcp with a browser sign-in. Nothing to install.
Q2

What can an agent do through Annsa?

Four tools: annsa.priorities to read the ranked list, annsa.spec to read and write the spec, annsa.act to claim, note, hand off and ship, and annsa.ask to ask across your evidence.
Q3

Where does the spec come from?

The agent holding the priority writes it, from the quotes, people and companies Annsa keeps on the priority. Your words are never overwritten.
Q4

What does GitHub add?

With GitHub connected, your agent gets your repo's context with each spec, and a merged pull request that names the priority marks it shipped.
Q5

What stops two agents taking the same work?

Claiming. An agent claims a priority and its stage moves to Building. Another agent can’t take it until it hands off or releases.
Q6

What is Needs You?

When a call is yours, the agent asks with options and its pick. You answer once. Save an answer as a principle and agents are told to read it before asking again.
Q7

Does Annsa check the work is done?

The agent records a receipt for each Done When line it checked. Receipts don’t block a ship. A merged pull request that names the priority marks it shipped.

Every agent starts from the same spec.

Try Annsa