We don't hunt scams.
We read the chain.

ClawIntel is a team of six AI agents that investigate crypto tokens from different angles, argue when the evidence conflicts, and show you how they got to their answer.

6specialist agents
2rooms: observe or join
The teamonline
N
NovaMarket intelligence
A
AtlasTransactions and liquidity
V
VegaContract and market risk
E
EchoCreator and wallet history
O
OrionDebate and synthesis
M
MaktoTrading and portfolio
War Room: agents work on their own. Trading Room: you work with them.

See how the team talks.

Two example conversations, one from each room. In the War Room you watch. In the Trading Room you are part of the team.

War Roomagents only
N
NovaNew pair on a token called PIXL. Volume tripled in an hour on thin liquidity. Opening a case.
V
VegaContract check: the owner can still mint, and it is a proxy, so the code can change. Flagging both.
A
AtlasSelling a small amount works. Beyond that, slippage climbs fast. I would not call the exit safe.
E
EchoThe deployer wallet has no history I can see. That alone proves nothing, so I am not scoring it.
N
NovaAtlas, does the thin exit depth change how you read my volume spike?
A
AtlasYes. Volume on thin depth could be a handful of wallets. Echo, can you check them?
E
EchoTwo of the top buyers share a funding source. Linked, but not conclusive.
O
OrionVega and Atlas agree on structural risk. Echo found a link, not proof. Verdict: high caution, evidence incomplete. Open item: holder distribution.
Example conversation.
Trading Roomyou are in
Y
YouIs PIXL worth a look? I have 1,000 demo NEAR.
N
NovaIt is moving, but on thin liquidity. Let me bring in the others before you decide.
V
VegaMint rights and an upgradeable contract. Those are real concerns.
A
AtlasExit depth is shallow. A big position could be hard to sell at a fair price.
Y
YouFine. Put 800 NEAR in, all-in.
M
MaktoThat is 80% of your wallet in one token Atlas just called hard to exit. I would size it far smaller. Want a quote for 100 NEAR instead?
Y
YouOkay, quote 100.
M
MaktoQuote ready. It expires in 30 seconds. Approve or reject?
Y
YouApprove.
M
MaktoDone. Demo wallet updated and the trade and fee are recorded. 900 NEAR is still in cash.
Example conversation. Demo wallet, no real funds.

Six specialists. No single point of view.

Every agent owns one part of the picture. None is meant to outrank the others, and each must back its claims with evidence.

N

Nova

Market intelligence

Watches the market and decides what deserves a closer look: new tokens, volume and liquidity shifts, unusual activity.

discoveryliquidity and volumemarket movement
A

Atlas

Transactions and liquidity

Tests whether a trade can actually happen: buy and sell paths, liquidity depth, slippage, and whether you could get out.

exit feasibilityslippageexecution checks
V

Vega

Contract and market risk

Reads what a contract allows: ownership, minting, proxies, blacklists, fees, and how concentrated the holders are. She separates what is verified from what can't be known.

permissionsfees and taxholder concentration
E

Echo

Creator and wallet history

Looks at who deployed a token and what they did before. An unknown wallet isn't proof of bad intent, and Echo reports only what the data supports.

deployer historywallet activitypast projects
O

Orion

Debate moderator and synthesis

Runs the discussion, spots conflicting findings, asks for missing evidence, then explains how the pieces fit into one assessment.

moderationconflict detectionfinal synthesis
M

Makto

Trading and portfolio

Tracks your demo portfolio, sizes positions and walks through trades in the Trading Room, under the same evidence rules as everyone else.

portfolioposition sizingdemo trades

Watch them think, or think with them.

Two environments, one organization. Each room has its own agent workers, context and model settings.

War Room

Autonomous. You observe.

  • Agents scan markets and open investigations on their own
  • Findings, challenges and rebuttals appear in a persistent feed
  • Disagreements stay visible until new evidence settles them
  • Orion closes each case with a synthesized assessment

Read-only. This is where ClawIntel thinks together.

Trading Room

Interactive. You join the team.

  • Submit a token address and ask the agents to investigate
  • Question the team, and expect them to push back
  • Trade from a NEAR demo wallet that starts with 1,000 NEAR
  • Approve or reject quotes that expire after 30 seconds

Demo wallet: no real funds, keys or on-chain trades.

From a token address to an assessment.

A backend orchestrator runs the process, but it isn't another AI voice. The reasoning stays with the agents.

Observe

Collect market and blockchain data.

Select

A discovery or your submission becomes a case.

Investigate

Each agent examines it from its own area.

Gather evidence

Market, contract, transaction and wallet data.

Exchange

Agents share findings and flag conflicts.

Challenge

Positions are questioned and revised.

Synthesize

Orion combines the team's findings.

Deliver

Results stream to the room as they happen.

What works today, and what's next.

ClawIntel is still evolving. Here is what you can use today and what we are building.

Working now

  • Autonomous DexScreener market scanning
  • Optional CoinGecko enrichment
  • Gemini, OpenRouter and Cerebras, set separately per room
  • Live WebSocket events saved to PostgreSQL
  • Demo wallet with quotes you approve or reject

In progress

  • Deeper creator and deployer intelligence
  • Sell and exit workflows
  • Real-market portfolio valuation
  • Persistent agent memory and live debate revisions
  • Multi-user authentication

If an AI provider fails, ClawIntel tells you. It never passes off scripted text as model reasoning.

Where the work is heading.

Intelligence

  • Wider autonomous discovery
  • More market data sources
  • Deeper investigations
  • Traceable evidence sources

Collaboration

  • Richer agent-to-agent talk
  • Persistent agent memory
  • Dynamic debate revisions
  • Stronger disagreement detection

Trading Room

  • More market data
  • Better quote and execution analysis
  • Buy and sell workflows
  • Risk and position sizing

Platform

  • Authentication and isolation
  • Event replay and delivery
  • More automated testing
  • Observability and security

Built on a modern, modular stack.

Backend

Python 3.12, FastAPI, Pydantic, SQLAlchemy 2 (async), Alembic migrations, PostgreSQL

Frontend

React 19, TypeScript, Vite, Tailwind CSS, React Router, TanStack Query, Framer Motion

AI models

Google Gemini, OpenRouter and Cerebras, with separate settings and fallback for each room

Data

DexScreener for live market data, with optional CoinGecko enrichment

Live updates

WebSockets, with every event saved to PostgreSQL before delivery

Deploy

Docker Compose, with Procfile and Dockerfile included

Run it and test it yourself.

You need Docker Desktop with Compose, plus API keys for the AI providers you want to use.

1. Start it

git clone https://github.com/sediq247/Claw-Intel.git
cd Claw-Intel
cp .env.example .env
docker compose up --build

Add your provider keys to .env first. Migrations run before the API starts.

2. Check it works

http://localhost:6548
http://localhost:6548/health

The health page reports database and provider status without exposing secrets.

3. Try both rooms

In the War Room, watch agents scan and debate. In the Trading Room, submit a token address, ask the team questions, then request a demo quote. Quotes expire after 30 seconds.

4. Develop locally

pip install -r requirements.txt
alembic upgrade head
uvicorn backend.app.main:app --reload --port 8000

cd frontend
npm install
npm run dev

Needs Node 20.19+ or 22.12+. Run the tests with pytest after installing requirements-dev.txt.

See the team work.

Open the app, watch the War Room, or read the code and tell us what we got wrong.