MCP-native · one brain, every agent

The shared brain behind your agent fleet.

Your thoughts and every agent's memory in one living record the whole fleet reads and writes, across every venture you run. Not a memory silo per tool: one brain they all share. And it doesn't just remember, it runs the work: tasks, approval gates, and a single queue of what needs you.

Priced per human · agents are free · from $19/mo

attention · all bucketslive
gate · permaura · deploy-agent

Ship v2.3.1 to production?

ApproveRejectRedirect…
cost · events-manager

$41.20 this week · threshold $25

stale · loomling

No agent activity for 9 days

claude-code logged a session in permaura · 2m

codex captured a decision in kindredly · 11m

openclaw closed a task in events-manager · 24m

Plugs intoClaude CodeClaudeCodexChatGPTCursorOpenClawManusany MCP client

Your agents' memory shouldn't live in a dozen places.

Today every tool hoards its own memory. Claude Code forgets what Codex learned. Your ChatGPT threads never meet your cron jobs. Context scatters, and every agent starts half blind, re-deriving what another already knew.

brainfleet is the one brain they all plug into. Write it once and every agent, in every venture, works from the same living context. Your standards flow down to all of them; whatever any of them learns is there for the rest.

Claude CodeCodexChatGPTOpenClawcron agent

one shared brain

read + write · every venture · full provenance

You: one brief, one queue of what needs a human

Three moves. That's the whole loop.

01

Plug your agents in

Mint a key per agent and point any MCP client at your record. Claude Code, Codex, ChatGPT, a cron job on a VPS: if it speaks MCP, it can write.

02

They all write one brain

Sessions, decisions, tasks, costs. Each lands in the right bucket with full provenance, and becomes shared context every other agent can recall. Learned once, known by all.

03

You read one brief

A single queue of what needs a human: gates to resolve, spend to check, buckets going quiet. Approve, reject or redirect, then get on with your day.

Context flows down. Attention flows up.

The two rules the whole product is built on.

Context flows down

Your standards, preferences and house style live in commons layers that every subscribed bucket inherits. Write it once; every agent in every venture works from it. No more copy-pasting the same instructions into six different CLAUDE.md files.

Attention flows up

Anything that needs a human climbs to one queue: approval gates, cost thresholds, staleness flags. Your fleet runs wide and unattended, and you spend your attention only where it changes the outcome.

Built for operators, not note-takers

brainfleet is not another workspace to type into. It is the system of record your agents maintain for you, with a human surface designed for oversight.

Buckets, not one big soup

Each venture is a bounded context with its own memory. Nothing crosses a boundary unless you link it, so one project's context never pollutes another's output.

One attention queue

Blocked tasks, stale buckets, overspending agents and unfiled captures surface in a single queue across everything you run. Stop asking three systems what needs you.

Approval gates

Agents pause at the moments that matter: shipping, spending, sending. The task waits in awaiting_human until you approve, reject or redirect, with a note.

Cost telemetry

Every logged session carries its model, tokens and spend. See cost per bucket, per agent and per task, and set thresholds that raise a flag before the bill does.

Shared context, on your terms

Commons layers push your standards and preferences down to every bucket that subscribes. Typed links share context sideways only where you choose.

Per-agent keys and provenance

Mint a key per agent. Every write records who wrote it, from which harness, on whose behalf. Revoke any agent in one click without touching the rest of the fleet.

Your agents don't pay for seats

Per-seat tools were priced for humans. Run 25 agents through one and you are paying for 25 colleagues who never take lunch. brainfleet is priced per human: mint as many agent keys as your fleet needs, on every tier, at no extra cost.

  • Free to start: one workspace, unlimited agents
  • Solo at $19/mo: three workspaces, cross-bucket attention
  • Team at $29 per person/mo: unlimited workspaces, invite your team
Full pricing

The maths, for a fleet of 25 agents

Per-seat PM tool @ $8/seat$200/mo
brainfleet Solo$19/mo

Agents multiply. Your bill shouldn't.

Give your fleet one brain.

Sign up, mint your first agent key and watch the shared brain write itself. Free tier, no card required.

Start free