Skip to content

Open source AI Employees. 8 scheduled business roles, 60 routines, on Claude Code and 10 other harnesses. They drive your browser the way you do and improve every run. You own the files.

3.6 KB
Raw Blame

Two guardrails, and what each routine does

The first guardrail is on outbound actions, and it is yours. Every AI Employee can send, post, submit, publish and spend. Shipped, every one of those is held: the draft is written, the form is filled and left open on the last step, the campaign arrives as a build sheet, and the last click is yours. RELEASES.md in the employee's folder is where you hand a channel over, one row at a time, with your conditions on it. From then on the routine that stages that channel completes the action itself and tells you in the morning brief what went out. Two of the eight ship with a channel already yours to configure: the SEO/AEO Employee publishes articles to the blog you name, and the Social Media Employee hands drafted posts to the channel you connect, after a veto window in the brief.

The second guardrail is on credentials, and it stays on. No AI Employee creates an account, enters or generates a password, completes a captcha, accepts terms, or writes a credential into any file. It never needs your password to do its job, so there is nothing to release.

On a hosted harness the guardrails are still yours to keep, and bots are not a boundary. On Grok Bot every bot on your account shares one cloud computer, one filesystem and every login on it. A release you write for one Employee is a file every bot can read, and a session you sync there is a session every bot can use. Scope by what you sign in to on that computer, never by which bot you talk to, and start read only on public pages before you sync anything.

Routine kind Reads Writes Leaves for you Holds, unless you release it
The standup, every weekday Every ledger, run record and tick since yesterday The board and the thirty line brief The brief, first thing Uses a browser
Sweeps (signals, prospects, inbox, site health, material, market) Your own signed in pages and public pages Dated, sourced ledger lines Nothing to do; the drafts come from these Types into a page, replies, marks anything read
Drafting (outreach, replies, posts, articles, creative) The ledgers, your strategy files, your proof inventory Queue files, unsent mailbox drafts, article drafts, creative sets The drafts, to edit and send Sends, posts, submits, or uses a number not in your proof inventory
Form filling (directories, press, listings) The card and your strategy files The queue entry with every field's value The tab, filled, on the last step Clicks Submit, Publish, Post, Send, Activate, Enable, or Create account
Account reads (ads, billing, analytics, registrar, host) Read screens in accounts you are signed in to Metrics ledgers, drift findings, build sheets Every change as a paste ready line with the screen named Changes a setting, saves a draft in an account that can spend
Weekly reviews and monthly refreshes A week or a month of the kit's own ledgers The scoreboard with a source beside every number, rewritten strategy where evidence disagrees One kill and one scale, as cards Estimates a number it did not measure
The Chief of Staff Every other AI Employee's run log, read only The fleet page, the fault dossier, the decision brief Three moves, argued both ways Writes into another AI Employee's folder

Everything the employee decides on its own lands as one dated line in a changelog you can read in a minute. Everything it learns about your sites lands in a recipe file inside its own folder. When it gets something wrong, one dated line in that file's ## Corrections section outranks the file from the next run on.