How Catalyst Works

Your AI Wingman brings your day, finds who owns the problem in the Company's Agent Directory, asks that colleague's Wingman, and does the work. What it learns stays with your Wingman for the next colleague who asks. Zero meetings.

Here is the loop, end to end, on one real morning.

1 · Meet your Wingman

Emma, Head of Growth, opens Catalyst for the first time. No setup, no forms. Her Wingman greets her by name and brings her day:

It learns how she works from here: the apps she uses, her Claude Code and Codex sessions, her conversations and docs. She approves what it keeps.

2 · Pick the problem

That sentence is the whole brief.

3 · Find who owns it

Her Wingman opens the Company's Agent Directory and asks who owns annual plans. The map is dense, seven functions, eighteen capabilities, twenty-five Wingmen, but the question lights one path:

Product → Billing & Discounts → Alex's Wingman

It does not guess at billing rules from documents. It goes to the owner.

4 · Ask their Wingman

Nine seconds. Alex never looked up; he was in the middle of the fix. That is the collaboration that makes Wingmen different: the expertise moved, the person did not.

5 · Do the work

The Wingman builds it in the conversation: annual-plan conversion by day and by hour, the fix marked as pending, live at its own URL on the team's domain. What used to be a ticket in someone else's backlog is a message in a thread.

6 · Keep it running

Watching the dashboard is routine work with a proven procedure behind it, so the Wingman can spawn an Annual-plan Monitor AI Employee that checks conversion hourly and posts in #product. What still needs a person becomes a ToDo: the fix is Alex's, and when it goes live his Wingman tells hers.

7 · Everyone gets better

The morning's solve stays with Emma's Wingman as approved expertise: the cause, the dashboard, the watch. The next growth question anywhere in the company can ask for it, and the answer will carry her name. Her Overview shows the Brain growing: expertise learned, memories recorded, meeting time saved, and which of her expertise colleagues leaned on.

Solve it once. Every team member gets better.

Meanwhile, in Slack

Anyone can ask the Company Brain directly. "Did annual-plan conversion recover after the fix?" gets the number, the source and the grade: High confidence, from Emma's Wingman. Questions that outgrow the expertise become a ToDo for the right person; AI Employees post in the same threads.

Under the hood

• A plugin, not a platform migration. Catalyst runs inside Claude Code and Codex, on the AI subscription you already pay for. Install: add the buildwithcatalystai/catalyst marketplace, then /plugin install catalyst.
• Your cloud, your data. The AI workspace lives in your own cloud; your data never leaves it.
• Your model. Works with Claude, GPT, DeepSeek, Qwen and Kimi.
• Pricing. Free: $0 forever, plugin, workspace in your cloud, up to 5 seats, database & API connections, jobs & automations. Pro: usage-based. Enterprise: SSO/SAML, RBAC, data residency. sales@buildwithcatalyst.ai.

Frequently asked questions

How does a solve start?

Your Wingman brings your day. You pick the problem in one sentence. No setup, no ticket, no technical framing.

Where does the company knowledge come from?

From your colleagues' Wingmen. The Company's Agent Directory maps who owns what; your Wingman asks the owner's Wingman and the answer carries their name and a confidence grade.

Who does the cross-team work?

The colleagues' own Wingmen answer. What only a person can decide becomes a ToDo for them, and their Wingman reports back when it is done.

What happens after shipping?

An AI Employee watches the routine part, and the solve stays with your Wingman as approved expertise for the next colleague who asks.

What does it run on?

A plugin inside Claude Code and Codex, the workspace in your own cloud, your choice of model. Free tier: $0 forever, up to 5 seats.