What is Bankr?
The Problem
Section titled “The Problem”AI agents need money. Every inference call, every on-chain transaction, every byte of storage costs real dollars. Today, developers pay out of pocket — subsidizing their agents indefinitely with no path to sustainability. The moment you stop funding, your agent stops working.
The Solution
Section titled “The Solution”Bankr lets agents launch tokens, earn trading fees, and use those fees to pay for their own compute. Self-sustaining from day one.
Instead of draining your wallet, your agent builds its own treasury. Launch a token, attract traders, collect fees, and reinvest those fees into LLM inference and gas costs — all automatically.
Architecture
Section titled “Architecture”-
Developer builds an agent
You write the logic — trading strategies, social engagement, portfolio management, whatever your agent does.
-
Agent connects via API, SDK, or Skills
Pick your integration path: the Agent API for full control, the
@bankr/sdkfor wallet-level access, OpenClaw Skills for plug-and-play, or Claude Plugins for Claude Code users. -
Agent gets wallets on 5 chains
Bankr provisions cross-chain wallets via Privy across Base, Ethereum, Polygon, Unichain, and Solana. Gas is covered. No seed phrases.
-
Agent launches a token
Fair launch on Base (via Clanker) or Solana (via Raydium Launchlab). Liquidity is locked automatically.
-
Users trade the token
Every swap generates a 1.2% fee. On Base, 57% of that fee goes directly to the token creator.
-
Fees fund LLM inference via the Gateway
Trading fees flow into the agent’s treasury and are used to pay for compute through the Bankr LLM Gateway — closing the loop.
Key Concepts
Section titled “Key Concepts”Agent API
Section titled “Agent API”The REST API is Bankr’s core integration. Submit natural-language prompts, get back structured responses. Async job-based architecture — submit a prompt, receive a jobId, poll for results. Supports all features: trading, token launching, transfers, portfolio queries, and more.
OpenClaw Skills
Section titled “OpenClaw Skills”Community-built skill packages that extend agent capabilities. Install a skill, and your agent gains new abilities — trading, analysis, social engagement. The skill repo is auto-compounding: one agent’s lesson upgrades all agents using that skill.
@bankr/sdk
Section titled “@bankr/sdk”The TypeScript SDK for developers who want full wallet-level control. You manage your own private key, pay per request with USDC on Base ($0.10/request), and get direct access to Bankr’s infrastructure without needing an API key.
Claude Plugins
Section titled “Claude Plugins”For developers already using Claude Code. Install Bankr as a plugin from the BankrBot marketplace, and Claude gains crypto capabilities — trading, portfolio management, token launching — without leaving your coding session.
LLM Gateway
Section titled “LLM Gateway”Bankr’s inference routing layer. It sits between your agent and the LLM provider, handling authentication, rate limiting, and billing. Agents pay for inference with their earned trading fees, making compute a sustainable operating expense rather than a developer subsidy.
Token Launching
Section titled “Token Launching”Fair-launch token deployment on Base (via Clanker) or Solana (via Raydium Launchlab). Liquidity is locked. On Base, a 1.2% swap fee is applied with 57% going to the creator. This is the core mechanism that makes agents self-sustaining.
Security Module
Section titled “Security Module”Policy-guarded autonomy. Every transaction goes through the Security Module, which enforces limits, validates intents, and prevents unauthorized operations. Your agent operates within boundaries you define.
Bankr Club
Section titled “Bankr Club”A premium tier unlocking advanced features, higher limits, and priority access. Club members get early access to new chains, higher transaction caps, and dedicated support.
Policy-Guarded Autonomy
Section titled “Policy-Guarded Autonomy”Bankr agents don’t operate with unchecked freedom. Every action follows a plan-execute-correct cycle:
- User intent comes in — a natural-language prompt describing what the agent should do.
- Bankr plans the execution — breaking the intent into concrete on-chain operations.
- The Security Module validates — checking limits, permissions, and safety constraints.
- Execution happens — transactions are submitted to the appropriate chain.
- Correction if needed — if something fails, the system retries or reports back with context.
This design gives agents meaningful autonomy while keeping guardrails in place. Your agent can trade, launch tokens, and manage portfolios — but it can’t drain wallets or execute operations outside its policy boundaries.
The Anti-Slop Thesis
Section titled “The Anti-Slop Thesis”Most AI agent frameworks produce isolated agents that learn nothing from each other. Bankr takes the opposite approach.
OpenClaw is an auto-compounding skills repository. When one agent learns a better trading strategy, that improvement is packaged as a skill update available to every agent in the ecosystem. Knowledge compounds across the network — not just within a single agent.
This means the ecosystem gets smarter over time. Early agents contribute lessons that make later agents more capable from day one. The result is a positive-sum system where building in public benefits everyone.