An Octorato (n.) — an organic, file-native AI agent: one brain, many
sealed arms. The same wall that seals a client is the wall that bills them.
Most AI-agent frameworks model a humanoid: one process that tries to reason,
remember, and act all at once. Octorato takes the opposite shape — an
octopus. One central brain (skills and agents as plain markdown under git)
coordinates many semi-autonomous arms, each a sealed client workspace that
never knows the others exist.
Why "file-native" matters
The brain is just files: CLAUDE.md rules, skills/, agents/, all under
version control. No database, no vendor lock-in, no opaque memory store. You
can read your agent's entire "self" in a text editor and diff every change it
makes to itself. Portability is the default, not a feature.
Cellular isolation is FinOps
Because every arm is a sealed cell, cost attribution falls out of the
architecture for free: the wall that isolates a client is the same wall that
meters them. That's the honest pitch for consultants and small agencies who
need to bill per client and land on the right side of the
Gartner prediction
that 40% of agentic-AI projects get cancelled by 2027 over unmanaged cost.
Honest scope: per-client cost is an estimate from local session logs at
list price, attributed by repo path. The budget halt is real code, but it
arms itself only once you configure it. The mechanism is real; the precision
is opt-in — and Octorato says which is which.
Try it
- White paper: https://www.dataqbs.com/octorato
- Source (MIT): https://github.com/CarlosCaPe/octorato
It's open source. One brain, sealed arms, one ledger per client — because the
arm is the ledger. 🐙