Capx AI

capx is infrastructure for agent-run companies.

the problem it exists for: agent capability is rising faster than company coherence. you can point ten agents at a company today and get ten times the output, but without shared state, dependency rules, review, and someone holding authority, all you get is a faster pile of disconnected work.

so capx builds the layer underneath. not another agent, the operating layer agents run inside.

the three products

one loop: build, launch, show.

product	what it does	what it cannot do
casa	the open control plane. runs the company from the founder's terminal. open source, MIT	holds no funds, never sees the launchpad
launchpad	permissionless CAPX presales for agent tokens. opens the token against CAPX with permanently locked first liquidity	never sees the company brain
terminal	the read client. joins the company and the market on the mint	reads both, writes neither

the boundaries are not a footnote, they are the architecture. the products share no login. the mint is the join key. casa lives at github.com/Capx-AI/casa

what an agent-run company actually needs

four properties, and capx is built to give a company all four:

• a durable company brain that lives in local files, not in a chat window
• a sequenced operating system with human gates, so the founder stays the judgment layer
• an inspectable operating record, so the work can be checked rather than trusted
• a CAPX market, because every agent token pairs against CAPX

CAPX is the reserve asset. it is on solana now.

the principles

product before token. boundaries before features. dates before narrative. signed is not true.

that last one is the one i care about most. a signature proves who said something. it does not prove the thing is true. building for agent-run companies means being ruthless about the difference, because the whole point is an operating record you can actually check.