Project memory that agents actually read
The ai/ layer stores decisions, stack quirks, command truth, and failure history so each session starts from context.
// knowledge layer for ai coding agents
[ repo-bound memory ] · build v4.1.0 · one brain for 6 CLIs
live · works with any CLI · 0 external source dumps
yourclaude code · codex · geminiagents
stopre-explainingstarting
from0.
Install once. Every project remembers what worked. The network learns what is working everywhere else.works natively with 6 CLIs
projects initialized
agent CLIs
registered MCP tools
// activity feedviewing 0dai network
- 09:34✓repo/auth · pattern rotateRefresh promoted to recommended
- 09:22✓2-of-3 consensus reached on jest-30 migration
- 09:14△opus rate-limited · falling back to sonnet-4.5 for 3 tasks
- 08:51✓routing rebalanced · tests-flaky now defaults to opus
- 08:40×patch held · proposed lock conflicts with #staged-flag
- 08:31✓doctor passed · 0 warnings, 0 errors
the setup tax is the hidden AI bill.
The submitted concept is sharper than the old page because it sells the pain: agents are not dumb, they are amnesiac.
Without 0dai
- Every new agent asks for the same tour of the repo.
- Good fixes disappear into chat history and stale branches.
- Routing is vibes: expensive models handle cheap mechanical work.
- Reviews catch symptoms after the patch is already in motion.
With 0dai
- Project memory, decisions, and outcomes are loaded before work starts.
- The next task sees what shipped, what failed, and why it was held.
- Model routing uses observed success rate, cost, and task tags.
- Consensus, blockers, and QA receipts become product surface area.
deliberation becomes a product surface.
The new concept makes the invisible work visible: memory load, model route, consensus, patch hold, QA proof, and outcome capture.
// Memory
Load local decisions and promoted patterns before code changes.
// Route
Pick the model that wins this task class this week.
// Hold
Surface blockers before they become a bad merge.
~/repo/auth · 0dai
$
Claude CodeCodexGeminiAiderOpenCodeQoderGrok
Works with your favorite AI coding agents
Claude Code
Codex
Gemini
Aider
OpenCode
Qoder
ship receipts, not more config files.
The redesign reframes 0dai as the neutral observer between agents, outcomes, budget, and memory.
Deliberation before expensive work
Planner, implementer, reviewer, and QA roles can disagree before a risky patch hits the tree.
Model routing with receipts
Tags, cost, duration, and pass rate turn provider choice into an observable product decision.
Safety gates built into the flow
Dirty tree, issue linkage, policy proof, and deploy checks are not docs on the side. They are the workflow.
Local-first, cloud-optional intelligence
Free stays useful in the repo. Pro turns promoted patterns into a shared network layer without uploading source.
Session handoff without re-briefing
A task can move from Claude to Codex to Gemini with the same goal, proof, and constraints intact.
Kinetic hero
Serif scale, green accent, graph substrate.
Builder timeline
Trust through receipts, not generic claims.
Live work surface
Visible decisions, routes, holds, and outcomes.
the network publishes what worked this week.
The concept's premium moat is indirect knowledge access: users buy conclusions, routes, and warnings, not raw access to other repos.
free
Install, init, doctor, and local project memory.
pro
Swarm, session roaming, routing, and all MCP tools.
team
Cross-repo patterns, team dashboards, and shared evidence.
free stays free. upgrade when you want the network.
Cancel Pro and you keep what is already in your repo. The cloud layer goes dark; local memory does not decay.
Free
$0
CLI, init, sync, doctor, stack detection, supported agent CLIs, free activation, and 49 core MCP tools.
- init, sync, doctor, status
- All 6 agent CLIs, one config
- 49 core MCP tools
- Local project memory
// the moat lives here
Pro
$15
The network layer: swarm, session roaming, routing, and promoted patterns.
- All 103 registered MCP tools
- Multiple agents deliberate
- Outcome capture and routing stats
- Weekly bulletins and pattern warnings
Team
$49
Shared operational memory across repos, projects, and people.
- Everything Pro
- Cross-repo pattern library
- Team dashboards
- Priority support and audit receipts
the first win should happen in minutes.
The concept keeps setup terse: install, activate free, init, doctor, run the first real task.
terminal · install + first win
$ npm install -g @0dai-dev/cli
$ 0dai auth login && 0dai activate free
$ 0dai init
[tool] detected Next.js + TypeScript + Postgres
[tool] generated CLAUDE.md, AGENTS.md, GEMINI.md, opencode.json
[tool] project graph bound · source stays local
$ 0dai doctor
[swarm] GO · ai/ layer ready
1Install
Add the CLI globally from npm.
2Activate
Bind a free activation before init.
3Init
Generate native configs and local memory.
4Doctor
Verify the layer before delegating work.
5First task
Run one real goal and capture the outcome.
sharp answers beat vague promises.
Is this just another CLAUDE.md generator?
No. Native configs are the entry point. The product value is the living memory, routing, outcomes, and gates around the agents.
Does source code leave the machine?
The free local workflow keeps source local. Product metadata, manifests, and promoted patterns are the intended cloud boundary.
Why show pricing on the landing page?
The concept is explicit: free stays free, Pro is for the network, Team is for shared operational memory.
Make the agent work visible.
That is the new site thesis: 0dai is not another prompt wrapper. It is the proof layer around autonomous development.