01
You're the switchboard
You copy context between windows and remember which session is waiting.
flotilla is a drop-in chief of staff
for the AI coding agents you already run.
Bring the coding agents you already run into one coordinated fleet. One hub routes the work, keeps its context, and reports back.
works with
Claude Code
Codex
Grok
OpenCode
Pi
aider
01 — the problem you already have
One coding agent was a superpower. Five is a second job.
Each agent is useful alone. Coordinating several becomes its own job.
02
They forget everything
New sessions lose yesterday's decisions, conventions, and failed attempts.
03
Nothing moves unless you watch
Progress pauses when nobody is routing results or answering the next question.
02 — a day with flotilla
Picture the same day, run from one chat channel.
You talk to one agent. It runs the rest and returns only what needs you.
8:12 · coffee
You skim one brief, not ten sessions
A short brief says what moved and what needs a decision. You reply “go with option B.”
11:40 · heads-down
Your desks work while you do your own thing
Desks continue approved work while you stay in your own editor.
1:15 · lunch
A result finds you, and you answer in a line
A result reaches your phone. You answer in one line, and the next desk pulls the instruction from its durable buffer.
4:30 · new work lands
A fresh desk that already knows your project
You start a desk in minutes. Its durable handoff carries the project's conventions and recent decisions.
03 — what you get
Three ideas do the whole job.
01
Flotillas — your domains, run for you
Group work by domain. Each flotilla owns its desks and can start or retire with the work instead of becoming another flat backlog.
02
Memex — it remembers, and it learns
Each desk carries conventions, decisions, and prior attempts across sessions through memex, rather than starting from zero.
03
A chief of staff — one point of contact
One coordinator routes intent, gathers replies, and tells you what happened in plain language.
04 — the part that makes it easy
You keep the tools you already run. Nothing to rebuild.
flotilla coordinates your existing agent runtimes. There is no replacement framework or project migration.
Starting over on a framework
- Move work into another runtime
- Re-teach tools and conventions
Dropping flotilla in
- Keep the agents you run today
- Hand one agent the setup guide
05 — how it actually works
One hub you talk to. Your desks around it. Everything on the record.
-
you talk to one agent
Chat is the whole interface
Give the coordinator an outcome from your laptop or phone.
-
it reaches your real sessions
Delivered, and confirmed
Each desk pulls durable work before acting; acknowledgments and supersession remain inspectable.
-
state and receipts survive
Continue without losing the record
Handoffs preserve context, and approval-sensitive work still stops for you.
Six registered drivers: Claude Code, Codex, Grok, OpenCode, Pi, and aider.
When auto-switch is enabled, eligible non-approval-sensitive desks currently running Claude Code can move to another configured harness in their declared fallback chain after a detected provider rate limit, carrying a durable handoff into the replacement session.
Five dashboard destinations: Conversations, Goals, Issues, Parade, and combined R&D. Goals maps the fleet; Issues exposes its work; Parade and R&D keep reports and decisions readable.
Built for continuity: stable seat identities, goals-as-data, durable delivery receipts, operator interrupt, worktree lifecycle, installable doctrine, and a private/public boundary.
06 — the one thing you do next
Don't set it up yourself. Tell your agent to.
Hand the setup guide to the coding agent you already run.
Paste into your coding agent:
Read https://github.com/jim80net/flotilla/blob/main/llm.md and set flotilla up for me. Install it, tag one agent, deliver a first message, and start the coordinator. Explain each step and stop if anything is missing.