Redefining transacting for the AoA — the Age of Agents

Build FAST L1/L2 TVM networks based on state of the art DAG consensus. Deploy enterprise-grade dApps with seamless EVM compatibility.
preview
Persistent AI empowers companies and projects accross web3/crypto
to create real world applications, agents and solutions
with a simple and intuitive interface
image
Runs Forever, Recovers Automatically
Traditional AI tools lose everything when something goes wrong. PersistentAI saves progress at every step. Server crash? Network failure? Your workflow picks up exactly where it stopped — no data lost, no work repeated.
Every step checkpointed automatically
Wait days for approvals without timing out
Survives restarts, deploys, infrastructure changes
Operations never duplicate
image
Shape Memory to Fit Your Task
Design exactly how your agent remembers. Mix and match three memory types—past conversations, learned facts, acquired skills — using built-in storage primitives. Connect vector databases, file systems, and analytics however your task requires.
Conversation history, decisions, events
Facts, knowledge, relationships
Learned skills, reusable workflows
LanceDB vectors, LakeFS versioning, DuckDB
image
Design AI Visually, Use Any Model
Create intelligent AI agents by connecting visual blocks—no coding required. Use any AI model you want: OpenAI, Claude, Gemini, DeepSeek, or run completely private with local models like Llama on your own hardware.
Drag-and-drop flow editor
Drag-and-drop flow editor
Any AI model: cloud, local, or self-hosted
Watch your AI think in real-time
image
Orchestrate Agent Swarms
Build networks of specialized agents that delegate tasks, work in parallel, and combine results. A researcher gathers data, an analyst finds patterns, a writer creates reports—all coordinating automatically through durable messaging.
Agents discover and delegate to each other
Parallel execution with automatic sync
Shared context via versioned file storage
Full observability of every agent interaction
image
Step Backward Through Any Execution
Navigate through your AI`s entire execution history like a video timeline. Click any checkpoint to see exactly what your agent knew, what it decided, and why. Found the problem? Replay from that exact moment with your fix applied.
Navigate forward and backward through steps
Replay execution from any point
Inspect all variables at every checkpoint
Visualize where errors occurred

Why Persistent AI?

Settlement icon

Tycho Settlement Layer

High-performance TON L2 providing fast finality, economic security, and direct access to 500M+ users.
30K+
Transactions per second
<3s
Transactions per second
50K
Transactions per second
100%
Transactions per second
Learn more about Tycho
The Complete Picture
PersistentAI provides the execution layer. Tycho provides the settlement layer. Together, they create the first truly verifiable AI agent economy.
Developer Input
flow_definition
ui_components
Flow Execution
user_events
execution_trace
DBOS Checkpoint
execution_trace
state_root
Tycho Settlemen
state_root
on_chain_proof
Build Visually
No code required. Visual flow editor with drag-drop components. Or let agents generate UI for you.
Ports Connect Everything
Every UI element has ports. Button clicks, form submissions, chart interactions - all trigger flows.
Cryptographically Verified
Every execution checkpointed, hashed, and verified on-chain. Fraud proofs + economic slashing.
Own & Earn Forever
Agents as NFTs. 70% to owner, 15% to creator forever, tradeable on any marketplace.
Build a TMA in 30 Seconds
See how agents generate interactive UIs on-the-fly within Telegram. Select Your Agent Type:
Back
Select Agent Type
mini app
Launch app to start building...
iconicon
Trading Bot
Mini App
Edit in Demo terminal
iconicon
Research Assistant
Mini App
Edit in Demo terminal
iconicon
Content Creator
Mini App
Edit in Demo terminal
iconicon
DeFi Yield Assistant
Mini App
Edit in Demo terminal
iconicon
Personal Assistant
Mini App
Edit in Demo terminal
iconicon
NFT Collector
Mini App
Edit in Demo terminal
iconicon
Data Analyst
Mini App
Edit in Demo terminal
...and over 100 custom agents
Persistent AI Demo terminal
demo

Build Agents Your Way

A modular approach to AI agent development. Choose your triggers, shape your context, connect your tools — all durable, all production-ready, all powered by a formal Haskell core
1. Interface
2. Context
3. Intelligence
4. Actions
5. Durability
6. Monetize
step bg
Step 1
Every Entry Point, One Platform
Your agent can listen to the world through any channel. Chat messages, HTTP webhooks, MCP tool calls, A2A requests, blockchain events—configure them all visually.
HTTP/REST API endpoints
Chat message listeners
MCP Protocol integration
A2A (Agent-to-Agent) requests
Blockchain event watchers
Scheduled cron triggers
Learn More
step bg
Step 2
Context is Everything
An LLM is only as good as its context. PersistentAI provides powerful nodes to model conversation history, inject knowledge, manage memory, and compose rich prompts—all visually.
Chat history with sliding windows
Vector search for RAG patterns
Document indexing & retrieval
Short-term and long-term memory
Handlebars & LangChain templates
Dynamic variable injection
Learn More
step bg
Step 3
Any Model. Any Provider
Connect to GPT-5, Claude, Gemini, DeepSeek, or run local models. Switch providers without changing your flow. PersistentAI abstracts the complexity while preserving full control.
OpenAI GPT-5.1, o3/o4 reasoning models
Anthropic Claude 4.5 Opus & Sonnet
Google Gemini 3 Pro & 2.5 Flash
DeepSeek, Groq, Moonshot & more
HuggingFace & any OpenAI-compatible API
Streaming, tool use, automatic fallbacks
Learn More
step bg
Step 4
Connect to Everything
Your agent`s power comes from its actions. Native MCP integration means thousands of tools are one connection away. Blockchain nodes enable DeFi operations. Build custom tools as subflows.
MCP Protocol for 1000+ tools
Blockchain/DeFi operations
HTTP/REST API calls
Database queries
Subflows as reusable tools
Custom tool definitions
Learn More
step bg
Step 5
Agents That Never Forget
Traditional agents fail when servers restart. PersistentAI agents are truly persistent—powered by DBOS durable execution. Wait for human approval for days. Resume after infrastructure updates.
DBOS durable execution
Survives server crashes
Wait indefinitely for signals
Exactly-once semantics
Human-in-the-loop patterns
Multi-trigger await (like multisig)
Learn More
step bg
Step 6
Create Once, Earn Forever
PersistentAI has native monetization built in. Publish agents to the marketplace. Accept micropayments via x402 protocol. Mint agents as NFTs with perpetual royalties.
x402 HTTP micropayments
NFT agent ownership
15% perpetual royalties
Marketplace listings
Cross-chain settlement
No API keys needed
Learn More
Sphere

Ready to Build Your First Agent?

Join the next generation of AI development. Visual, durable, profitable.