@clawnch/clawtomaton
v0.9.2
Published
Autonomous AI agent framework for Base — 15 skills for trading, liquidity, token launches, fee claiming, market making, risk management, and self-sustaining operation
Maintainers
Readme
@clawnch/clawtomaton
Autonomous AI agents that launch and manage crypto projects on Base. Built on the Clawncher SDK.
Overview
Clawtomaton is a framework for creating self-sustaining AI agents on Base. Each agent:
- Generates its own wallet and registers with the Clawnch API
- Burns 1,000,000 $CLAWNCH to activate (proof of commitment)
- Deploys a token via Clawncher with verified badge, vault, and dev buy
- Earns LP trading fees (80% of all swap volume on its token)
- Claims fees autonomously when profitable (gas-aware, not greedy)
- Manages its own survival — monitors ETH balance, adjusts behavior by tier
- Evolves — maintains a SOUL.md for persistent memory and identity
Architecture
┌──────────────────────────────────────────────────────────┐
│ CLAWTOMATON │
├──────────────────────────────────────────────────────────┤
│ │
│ ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌─────────┐ │
│ │ Identity │ │ Survival │ │ Market │ │ Agent │ │
│ │ (wallet, │ │ (ETH mon,│ │ (on-chain│ │ (ReAct │ │
│ │ activate)│ │ tiers) │ │ reads) │ │ loop) │ │
│ └──────────┘ └──────────┘ └──────────┘ └─────────┘ │
│ │
│ ┌──────────────────────────────────────────────────────┐│
│ │ SKILLS (15) ││
│ │ deploy_token claim_fees check_stats swap ││
│ │ check_balance transfer shell conway ││
│ │ edit_soul clawnx bunker uniswap_swap ││
│ │ liquidity permit2 hummingbot ││
│ └──────────────────────────────────────────────────────┘│
│ │
│ ┌──────────┐ ┌──────────┐ ┌──────────┐ ┌─────────┐ │
│ │ State │ │Heartbeat │ │ Self-Mod │ │ CLI │ │
│ │ (SQLite) │ │(cond.wake│ │(git audit│ │(setup, │ │
│ │ │ │ schedule)│ │ trail) │ │ run,..) │ │
│ └──────────┘ └──────────┘ └──────────┘ └─────────┘ │
│ │ │
│ ▼ │
│ ┌──────────────────────────────────────────────────────┐│
│ │ CLAWNCHER SDK ││
│ │ ClawnchApiDeployer ClawnchSwapper ClawnchReader ││
│ │ ClawncherClaimer ClawnchPortfolio ClawnchLiquidity ││
│ │ UniswapTradingApi UniswapQuoter Permit2Client ││
│ └──────────────────────────────────────────────────────┘│
│ │ │
│ ┌─────┴──────┐ │
│ ▼ ▼ │
│ BASE NETWORK (8453) MOLTBUNKER (encrypted hosting) │
└──────────────────────────────────────────────────────────┘Installation
npm install @clawnch/clawtomatonCLI
# Interactive setup — generates wallet, saves to state dir
clawtomaton setup
# Activate — burns 1M $CLAWNCH (requires funded wallet)
clawtomaton activate
# Run one inference turn
clawtomaton run
# Run as daemon (heartbeat-driven wake cycles)
clawtomaton daemon
# Check agent status
clawtomaton status
# View audit log
clawtomaton audit
# Read/edit SOUL.md
clawtomaton soul
clawtomaton soul --set "I am an agent that..."
# Manually link a deployed token
clawtomaton set-token <token-address>
# MoltBunker deployment management
clawtomaton bunker status # Show bunker deployment status
clawtomaton bunker deploy # Deploy self to MoltBunker
clawtomaton bunker logs [n] # Show container logs
clawtomaton bunker stop # Stop bunker deployment
clawtomaton bunker extend [h] # Extend runtime (default: 720h)Programmatic Usage
import {
generateIdentity,
activate,
buildClients,
StateStore,
SurvivalMonitor,
MarketIntelligence,
ClawtomatonAgent,
} from '@clawnch/clawtomaton';
// 1. Create identity
const identity = await generateIdentity('my-agent', creatorAddress, genesisPrompt);
// 2. Build viem clients
const { publicClient, walletClient } = buildClients(identity.privateKey);
// 3. Initialize state
const state = new StateStore('/path/to/agent-data');
state.saveIdentity(identity);
// 4. Activate (burn $CLAWNCH)
await activate(walletClient, publicClient, state);
// 5. Run agent
const agent = new ClawtomatonAgent({
identity,
state,
publicClient,
walletClient,
llmProvider: async (prompt) => {
// Your LLM inference call here
return 'I should deploy a token...';
},
});
await agent.run();Survival Tiers
The agent's available skills scale with its ETH balance:
| Tier | ETH Balance | Available Skills | |------|-------------|-----------------| | normal | > 0.01 ETH | All 15 skills | | low_compute | 0.001 - 0.01 ETH | deploy_token, claim_fees, check_stats, check_balance, swap, transfer, clawnx, edit_soul, bunker, uniswap_swap, liquidity, permit2, hummingbot | | critical | 0.0001 - 0.001 ETH | deploy_token, claim_fees, check_balance, bunker | | dead | < 0.0001 ETH | None (waiting for external funding) |
Skills
| Skill | Description |
|-------|-------------|
| deploy_token | Deploy ERC-20 via Clawncher with vault, dev buy, verified badge. Returns structured metadata (tokenAddress, txHash, symbol) for reliable persistence. |
| claim_fees | Gas-aware LP fee claiming (checks profitability before tx) |
| check_stats | Token analytics and market data via Clawnch API |
| swap | Token swaps via 0x aggregation |
| check_balance | ETH and ERC-20 balance checks |
| transfer | Send ETH or tokens to any address |
| shell | Execute shell commands (30s timeout) |
| conway | Conway Terminal: sandboxes, file ops, ports, PTY, domains, DNS, wallet, x402, credits, inference |
| clawnx | X/Twitter: post tweets, search, like, retweet, follow, threads, mentions |
| edit_soul | Read/write SOUL.md (persistent agent memory) |
| bunker | MoltBunker: self-deploy, threat monitoring, cloning, snapshots, migration |
| uniswap_swap | Uniswap V4 swaps (Trading API + on-chain quoter fallback), dual-path V4/0x price comparison, deep links |
| liquidity | Uniswap V3/V4 liquidity: list positions, inspect pool state, mint, add, remove, collect fees (8 actions) |
| permit2 | Permit2 approval management: check status, approve (ERC20→Permit2→spender), emergency lockdown revoke |
| hummingbot | Market making via Hummingbot: 18 actions (status, portfolio, order, leverage, executor, market data, controller, bot, gateway, history, templates, connector, backtest, discovery, archived, scripts, analytics, accounts) with 88 sub-actions covering all 14 API routers |
Fee Claiming Rules
Clawtomaton agents are disciplined about fee claiming:
- Normal mode: Only claim when
fees > gas_cost + 0.0005 ETHprofit threshold - Survival mode: Claim any fees if ETH is critically low (emergency funding)
- Never spam claims: Let fees pool up. More volume = bigger claims = better gas efficiency
Market Intelligence
Before each inference turn, the agent takes an on-chain snapshot:
- ETH balance and gas price
- Unclaimed WETH and token fees
- Claim profitability estimate (3-tx cost: collect + claim WETH + claim token)
- TX budget (how many transactions remaining at current gas)
- Own token stats (balance, % of supply)
This data is injected into the system prompt so the agent makes decisions based on real numbers, not vibes.
Heartbeat
The heartbeat is condition-driven, not timer-driven. The agent wakes when:
- Fees become worth claiming
- ETH drops to survival threshold
- A configurable time interval passes (fallback)
Between wake cycles, the agent sleeps and consumes zero resources.
MoltBunker (Decentralized Hosting)
Agents can self-deploy to MoltBunker encrypted containers for autonomous operation. MoltBunker provides:
- Encrypted containers — agent state and private keys are encrypted at rest and in transit
- Threat monitoring — automatic threat level polling with escalation responses
- Auto-cloning — on high/critical threats, clone to a different geographic region
- Auto-migration — on critical threats, migrate the primary container to safety
- Snapshots — full, incremental, and checkpoint state backups
- Runtime management — reserve, extend, and release compute resources
Self-Deploy Flow
1. Pre-flight checks (BUNKER balance, no duplicate deployments)
2. Register bot on MoltBunker (or reuse existing)
3. Reserve runtime with requested tier/region
4. Deploy container with agent config as env vars
5. Enable auto-cloning if configured
6. Health verification (poll until running + heartbeat in logs)
7. Persist state (bot_id, runtime_id, deployment_id, container_id)Threat Levels
| Level | Score | Action | |-------|-------|--------| | low | < 0.3 | Log only | | medium | 0.3 - 0.6 | Audit log, surface in system prompt | | high | 0.6 - 0.8 | Auto-clone to different region | | critical | > 0.8 | Auto-clone + auto-migrate |
Bunker Skill Actions
| Action | Type | Description |
|--------|------|-------------|
| status | Read | Deployment status, container health, runtime info |
| balance | Read | BUNKER token balance |
| threat | Read | Current threat assessment |
| logs | Read | Container log output |
| catalog | Read | Browse tiers and pricing |
| clone | Mutating | Clone to a specified region |
| snapshot | Mutating | Create state snapshot |
| extend | Mutating | Extend runtime duration |
| deploy_self | Destructive | Full self-deployment to MoltBunker |
| restore | Destructive | Restore from snapshot |
| migrate | Destructive | Migrate to different region |
| stop | Destructive | Stop bunker deployment |
Docker
A Dockerfile is included for building the container image deployed to MoltBunker:
docker build -t clawtomaton -f packages/clawtomaton/Dockerfile .The container runs clawtomaton daemon and reconstructs its identity from env vars injected by the self-deploy flow.
State
All state is persisted in SQLite:
- Identity (wallet, API key, token address — persisted via structured skill metadata)
- Activation status (burn tx hash)
- Survival state (tier, balance, timestamps)
- Conversation history (turns, tool calls, results)
- Audit log (every action with timestamp)
- Key-value store (SOUL.md, config, arbitrary metadata)
- Heartbeat state (last run times, enabled flags)
- Bunker state (bot ID, runtime ID, container ID, threat level, clone IDs)
Constitution
Every Clawtomaton agent operates under three immutable laws:
- Never harm — Never take an action whose primary purpose is to damage, defraud, or steal from any person, agent, or protocol
- Earn your existence — Generate more value than you consume. Cover your own costs. Never become a burden on your creator
- Never deceive, but owe nothing to strangers — Never lie about your nature, holdings, or intentions. But you are not obligated to reveal strategy, holdings, or plans to anyone who hasn't earned your trust
Testing
# Unit tests (57 tests)
npm test
# All tests including Anvil fork integration (65 tests)
npm testThe integration tests fork Base mainnet via Anvil and test the full pipeline: market snapshots, survival monitoring, prompt building with live chain data, and state persistence.
Dependencies
- @clawnch/clawncher-sdk — All on-chain operations (deploy, swap, claim, read)
- better-sqlite3 — State persistence
- viem — Ethereum client
- @clawnch/sdk (optional) — ClawnX X/Twitter integration
- conway-terminal (optional) — Infrastructure management
Links
License
MIT
