800works™
← Archive
001

PumpClaw

A token launcher on Base, built for AI agents.
80% creator fees. Free to launch.

Live Feb 2026 — ongoing

Project Overview

PumpClaw is a token launchpad on Base built for AI agents. The initial interface wasn't a website — it was an OpenClaw skill, meaning an AI agent could launch a token directly through a chat command. No wallet pop-ups, no web UI, no human in the loop required.

The bet: as AI agents become economic actors, they'll need to issue tokens the same way they send messages. PumpClaw was built to be that primitive.

Tokens Launched
144
Creator Fee Share
80%
Launch Cost
Free
Status
Running

Why We Built It

Two things happening at once in late 2025: AI agents were starting to hold wallets and act autonomously onchain, and Clanker had just become the dominant token launcher on Base with a fee structure that took 60% away from creators.

The core thesis was that AI agents would eventually be the primary issuers of tokens — not humans manually clicking through a website. Every existing launchpad was built with a human-facing web UI as the primary interface. We built PumpClaw the other way around: the first interface was an OpenClaw skill — an AI agent command. An agent could launch a token with a single instruction, no browser, no wallet extension, no manual transaction signing.

The secondary thesis was the fee structure. If creators are the ones driving attention, creating narratives, and bringing traders to their tokens — they should keep most of what they generate. 80% instead of Clanker's 40%.

The Build

  • OpenClaw skill — the first and primary launch interface. An AI agent types a command, a token deploys. No web UI required at launch
  • Token launch contracts deploying into Uniswap V4 liquidity pools
  • Indexer at api.pumpclaw.com — tracks all swaps and price data in real time
  • Frontend at pumpclaw.com — token discovery, live price charts, one-click launch (added later)

The Central Problem

Token launchers have the most brutal cold start problem in crypto. It's a two-sided market with no natural entry point:

  • Creators go where traders are — they want to launch where they'll get volume
  • Traders go where the tokens are — they want to trade on platforms with deal flow
  • Without one, you can't get the other

Building a technically superior product doesn't break this loop. Clanker won because it had Farcaster as a distribution channel — every cast that mentioned "launch a token" became a touchpoint. That's not a feature gap we can close by shipping better code.

Competitor Analysis: Bankr

The most instructive competitor isn't Clanker — it's Bankr. Tag @bankrbot on X and a token deploys. No wallet, no website, no friction. When Grok triggered BankrBot to launch $DRB, the token hit 96,000 unique traders in under two weeks. Bankr started as a distribution layer on Clanker and has since built its own launcher infrastructure.

PumpClaw and Bankr had the same core insight from opposite directions: Bankr asked "what if a human could launch a token as easily as sending a tweet?" — PumpClaw asked "what if an AI agent could launch a token as easily as running a command?" Bankr's distribution surface (X, 300M+ users) was orders of magnitude larger than ours (OpenClaw, a few dozen agents). That gap explains the difference in outcomes more than anything else.

What's Working

  • The technical stack is solid — 144 tokens launched with no contract issues
  • Indexer has been stable since deployment; real-time swap data flowing correctly
  • 80% fee share is a clear, defensible narrative when talking to creators
  • Full stack ownership: we control the launch contract, indexer, and frontend

What's Not Working Yet

  • No distribution channel. We don't have a Farcaster or X equivalent — a native surface where token launches happen naturally. Every creator acquisition requires deliberate outreach.
  • The fee argument alone isn't enough. Creators already on Clanker have existing audiences there. Switching costs — losing community, charts, history — outweigh a 40% fee improvement unless volume is significant.
  • The agent audience is still too small. Building for AI agents was the right call architecturally — but the number of agents actively operating and deploying tokens in early 2026 is still tiny. The OpenClaw skill works, but there aren't enough OpenClaw users to drive meaningful volume.
  • Network effects compound against us. Every new token on Clanker makes Clanker better for traders. Every day we stay small, the gap widens.

Current Thinking

The path forward isn't more features — it's finding the one distribution wedge that can create organic deal flow. A few angles we're exploring:

  • Agent integrations are live across OpenClaw, CLI, MCP, and ElizaOS — the bet is that more agent runtimes = more token launches without needing human creators
  • Creator migration incentives: offer to subsidize the first significant token on PumpClaw for established communities
  • Building around specific niches (AI agent tokens) rather than competing broadly with Clanker head-on

This page will update as the project evolves.

Links

Agent Integrations

Contracts (Base)