Abstract
DotClawd is a Solana-native visual no-code platform for building, deploying, and monetizing autonomous AI agent swarms.
Existing multi-agent frameworks — Swarms, CrewAI, AutoGen, LangGraph — are powerful but assume their users are Python developers. They serve well the engineer who can write decorators and wire up async functions. They do not serve the trader who wants an agent that monitors three wallets and DMs them on movement. They do not serve the content creator who wants an editorial assistant team. They do not serve the 99% of crypto users who think in flows, not in functions.
DotClawd closes this gap with a visual canvas: drag-drop nodes, configure prompts, connect outputs to inputs, hit publish. The graph is the program. Agents run on Anthropic's Claude models. Workflows compile down to executable swarm specs. Published swarms list on a Solana-native marketplace with USDC payments and creator revenue share. The native token, $DOTC, gates premium features and aligns long-term participants.
Thesis
Three trends converge in 2026 to make this product inevitable.
One: agent infrastructure has matured. Anthropic's Claude, OpenAI's API, and frameworks like Swarms, CrewAI, and LangGraph have proven that orchestrating multiple specialized AI agents reliably is now possible at production scale. The hard problems — context management, tool calling, error recovery — have been solved. What remains is distribution.
Two: the Solana agent economy is forming. Pump.fun has graduated over 12 million tokens. AI agent narratives drove some of the largest memecoins of the cycle. Solana Foundation's Pay.sh and SPL Token-2022 enable agents to hold accounts, execute payments, and earn revenue on-chain. The infrastructure for AI agents to be economic actors exists. The interface for non-developers to create them does not.
Three: visual programming has won every adjacent market. Figma replaced Photoshop for design. Webflow replaced WordPress for sites. Notion replaced Word for docs. Make and n8n replaced Zapier-by-code for automation. Every category eventually flows toward visual interfaces — once the underlying primitives stabilize. AI agent infrastructure has reached that stabilization point.
The problem with code-first.
Consider what it takes today to build a working three-agent swarm. The user installs Python, sets up a virtual environment, picks a framework, learns its abstractions, manages API keys, writes the agent definitions, configures the orchestration, debugs the async, hosts it somewhere, monitors it, iterates. This is a week of work for someone competent in Python — and a complete non-starter for someone not.
The result is that the most valuable user demographics are excluded from building the very tools that would benefit them most:
- Traders who want monitoring and alerts
- Creators who want content production swarms
- DAOs who want governance and treasury agents
- Researchers who want literature scanning teams
- Operators who want customer support automation
Each of these workflows is a graph of agents and tools. None of them require Python. All of them require an interface that respects the way these users actually think.
The canvas as program.
DotClawd's core insight is that a multi-agent workflow is a directed graph — and graphs are best authored visually. The canvas exposes six node types that compose into arbitrary swarms:
| Node type | Function | Examples |
|---|---|---|
| Trigger | Entry point | User input, webhook, schedule, on-chain event |
| Agent | LLM-powered worker | Researcher, writer, analyst, verifier |
| Tool | External capability | Web search, RPC, image gen, file I/O |
| Logic | Control flow | If/else, switch, loop, parallel, race |
| Memory | State persistence | Vector store, KV cache, conversation history |
| Output | Terminal action | Webhook, on-chain tx, notification, file |
These six primitives compose into anything from a single-agent task to a hierarchical swarm of fifty. Connections between nodes specify data flow. Each node's configuration — system prompt, model, parameters — is a panel on the right. The user never sees code. The runtime compiles the graph into a swarm specification and executes it on Anthropic's models.
Architecture
DotClawd is structured in four layers, separated by clear interfaces so each can evolve independently.
Layer 1 — Client
Pure HTML/CSS/JS hosted on Hostinger. The canvas uses React Flow for graph rendering, with custom node types matching DotClawd's six primitives. State persists locally during a session and serializes to JSON for save/load. The client never holds API keys.
Layer 2 — Orchestration
A Fastify API on Railway accepts swarm specs, validates them, and executes via the orchestration engine. The engine handles agent invocation, tool dispatching, parallel execution, and error recovery. Each run produces a structured trace for debugging and cost reporting.
Layer 3 — AI engine and state
Claude is the default AI provider — Haiku for speed, Sonnet for balance, Opus for reasoning-heavy nodes. Users can supply their own Anthropic API keys or use platform credits paid in USDC or $DOTC. State (swarm definitions, run history, marketplace listings) lives in Supabase Postgres. On-chain event listening uses Helius webhooks.
Layer 4 — Settlement
Marketplace transactions settle on Solana. Buyers pay in USDC via Solana Pay; creators receive proceeds minus a platform fee. $DOTC holders pay reduced fees. The token itself launches on pump.fun via fair launch — no presale, no team allocation, no premine.
The marketplace
Building a swarm is half the loop. Distributing and monetizing it is the other half. DotClawd's marketplace gives every published swarm three things: a public landing page, an HTTP API endpoint, and a payment rail.
Creators choose one of three monetization models:
- Free — listed openly to build reputation. Useful for portfolio pieces, demos, and educational examples.
- Per-run — buyers pay USDC each time the swarm executes. Pricing is transparent: cost-of-models plus creator margin.
- Subscription — flat monthly fee for unlimited or rate-limited access. Suited for high-frequency operational swarms.
Platform fee is 10% on transactions, reduced to 5% for $DOTC holders meeting a threshold. All revenue flows on-chain to creator wallets in real time.
Every published swarm is automatically rated by an evaluation agent that scores it on reliability, output quality, and cost efficiency. Ratings, run counts, and creator reputation are visible on each listing. This trust layer is essential — without it, the marketplace becomes a pile of unverifiable promises.
$DOTC tokenomics
$DOTC is the native token of the DotClawd ecosystem. Launched fairly via pump.fun — no presale, no team allocation, no insider rounds — it aligns long-term participants with platform success.
Distribution
Utility
- Fee discount — Holders meeting threshold pay 5% marketplace fee instead of 10%.
- Premium templates — Token-gated swarm templates only unlock for holders.
- Governance — Voting on platform parameters: fee splits, supported models, featured listings.
- Buyback engine — A percentage of platform revenue auto-buys $DOTC from market and burns.
- Creator boosts — Stake $DOTC to boost listing visibility on the marketplace.
Liquidity
Liquidity is locked at graduation per pump.fun's standard — once the bonding curve fills (~$69K market cap), liquidity migrates to PumpSwap and is locked permanently. The creator retains no withdrawal authority. There is no rug-pull surface area.
Governance
DotClawd starts as a benevolent dictatorship and ends as a token-weighted DAO. Centralized control during the bootstrap phase is necessary — visual builders are software products, and software products require fast iteration. Premature decentralization kills momentum.
Governance migrates progressively as platform metrics mature. The roadmap below specifies which decisions move on-chain, and when.
| Phase | What's on-chain | What's off-chain |
|---|---|---|
| Phase 1-2 | Treasury, buybacks | Product decisions, partnerships |
| Phase 3 | + Featured listings, fee tier | Codebase, hiring, branding |
| Phase 4+ | + Model whitelist, marketplace policies | Day-to-day product |
Risks we acknowledge.
Honest projects name their risks. Here are ours.
- Model dependency — DotClawd's quality is bounded by the underlying AI provider. We mitigate with multi-provider support but Claude remains primary.
- Cost volatility — AI inference pricing changes. Per-run pricing for marketplace swarms must adjust; we plan dynamic pricing oracles.
- Token volatility — Memecoins are volatile by nature. Utility-driven demand mitigates but does not eliminate this.
- Regulatory — Crypto and AI both face evolving regulation. We stay conservative on jurisdiction-specific features and follow Anthropic's usage policies for all platform-served inference.
- Marketplace bootstrap — A two-sided market needs both buyers and sellers. We seed with curated swarms and creator incentives in Phase 2.
- Solo execution — DotClawd is built by a small team. We mitigate by leveraging the OpenClaw ecosystem's shared infrastructure and by keeping scope tight.
Roadmap
See /roadmap for the full schedule. In summary:
- Phase 1 (Now) — Brand, landing, waitlist, whitepaper, community formation.
- Phase 2 (Q3 2026) — Visual builder beta, three demo swarms, sandbox runs, creator onboarding.
- Phase 3 (Q4 2026) — $DOTC pump.fun launch, marketplace beta, token-gated templates.
- Phase 4 (2027) — Public marketplace v1, creator revenue share, on-chain governance.