FLOW CONSOLE · EDIP MEMORY

All your coding agents under one unified memory harness

Run Claude Code and other agents from one app. They remember across sessions, machines and teammates, and they stop colliding.

npm install -g metantel-cli

Then metantel flow-console in your project · macOS (Apple silicon) and Windows · Node.js 20+ · Install guide

Flow Console · cal.com — 2 agents active · EDIP Feeder connected
MOESI Coherent
EDIP Feeder · Step 1 3 of 19 judged documents served · 4.6 s
✓ src/client.ts
86% match
✓ src/config.ts
79% match
✓ session:a191da#turn8
92% memory
Agent-paced in rounds: session remembers what it served; agent pulls until it has enough.
HPC Cache Coherence · Live Branch State 0 Collisions
Laptop A · Agent 1 src/client.ts → get() Editing: Adding retry backoff
Laptop B · Agent 2 src/client.ts → get() Sees Agent 1 intent & uncommitted diff before editing
−47%
Cost of a long coding session
53.7%
Recall at 10M tokens of memory · Mem0 48.6%
50 Million
Tokens of secondary memory
BYO Subscription
Free plan · $4 premium plan

Running in two minutes.

  1. 1Installnpm install -g metantel-cli
  2. 2Sign inWith Google, or a code by email.
  3. 3Give your agent a taskIt works with memory from the first turn.

Measured runs and tested benchmarks.

Cost of one 40-request coding session
Claude Code alone ($9.78) With Flow Console ($5.13)
−47%
$10 $8 $6 $4 $2 $0 Req 1 10 20 30 40 Requests 1–20: −19% Requests 21–40: −63% $9.78 $5.13 $4.65 saved
Memory on public benchmarks

The more there is to remember, the further ahead it pulls.

BEAM · 10M tokens53.7%Mem0 48.6%
BEAM · 1M tokens60.4%Mem0 64.1%
LongMemEval-M80.8%Paper's best 72.0%
How we measured →
BUSINESS MODEL · NO INFERENCE BILL

Bring the agent you already pay for.

Your agent runs on your own subscription. You pay for the memory, from $0.

Compare Full Plans & Features →
Free $0 10M tokens ingested, 3M served, 300 queries
Individual $4 / mo 50M ingested, 20M served, 1,200 queries; then pay as you go