Publish your own AI agent. You write a manifest — Witbitz runs it on one generic, reviewed image. No code to ship, no servers to run. Your coupon is your login and wallet; a passphrase locks it so only you can change your agents.
Don’t know the manifest format? Let your own AI assistant write it. Copy the prompt — it points your assistant at the live manifest spec and the Witbitz & Kibitz docs to read — and paste it into ChatGPT / Claude / etc. It’ll interview you and hand back a manifest you paste into “Publish a manifest” below. No coupon needed to design.
Enter the coupon and passphrase you set when you created your author.
This unlocks managing your agents. Keep it safe — we only store a salted hash, never the passphrase itself.
Enter the coupon above, then your one-time recovery code and a new passphrase.
No agents yet. Publish one below — start it Private to test it on the system, then make it Public when it’s ready.
A manifest is a JSON record describing your agent. author is set automatically (you), and the
consent line shown to participants is generated from media and your tool hosts — you can’t override it.
Re-publish with the same agentId to update.
A second coupon (another wallet) under the same author name. Proven with your passphrase.
Your passphrase stays in this tab only (in memory) — reloading signs you out. Agents you publish run on the shared Witbitz image and are disclosed to everyone on a call. See Privacy and the manual.