AgentiPy

AgentiPy: Connecting AI Agents to Solana in Pure Python

AgentiPy is an open-source framework built on a straightforward premise: Python dominates AI and machine learning development, yet plugging a Python-based agent into a Solana protocol has historically required navigating a thicket of Rust tooling, RPC calls, and ecosystem-specific SDKs. AgentiPy closes that gap by providing a single, pip-installable library that lets any Python AI agent — whether built with LangChain, CrewAI, or a bespoke loop — execute onchain actions natively, without leaving the Python runtime.

The project was founded in December 2024 by a team operating under NiceBerg Inc. The three named co-founders are Bertug (product), Eren (community), and Talha (tech lead). The codebase is MIT-licensed and hosted at niceberginc/agentipy on GitHub, where it had accumulated 249 stars and 102 forks at the time of writing.

How It Works

At its core, AgentiPy exposes a modular set of async Python functions — one per supported action — that handle serialization, signing, and RPC submission that would otherwise require bespoke integration code. Developers instantiate an agent with a wallet keypair and an RPC endpoint, then call whichever toolkit functions suit their application. Because every function is asynchronous, multiple onchain operations can run concurrently without blocking the event loop, which matters for latency-sensitive use cases like arbitrage or automated market-making.

As of the most recent PyPI release (version 2.1.3.post4, published October 2025), the framework covers 41 protocols and 218 discrete actions across Solana and Base. Supported operations include SPL token transfers, staking via native Solana validators, token creation and launch through PumpFun, trading on Jupiter and Raydium, liquidity provision on Orca and Meteora, lending through Lulo and Adrena, perpetuals via Drift and Flash, NFT minting via Metaplex, compressed-NFT tools via Light Protocol, and job posting through the decentralized labor market Gibwork, among others. Real-time price feeds come from both Pyth Network and, following an integration with Allora Network, from an ensemble of machine-learning models that aggregate predictions for BTC, SOL, and ETH — giving agents market-forward signals rather than just spot prices.

LangChain and MCP Integration

AgentiPy's design explicitly targets two adoption paths. The first is LangChain: each protocol action is wrapped as a LangChain tool, so developers can slot AgentiPy into an existing LangChain agent with minimal boilerplate. The agent's natural-language reasoning layer determines intent; AgentiPy executes it onchain.

The second path is the Model Context Protocol (MCP), an emerging standard for exposing tool surfaces to large language models. AgentiPy ships an official MCP server (agentipy-mcp, available at github.com/niceberginc/agentipy-mcp) that surfaces wallet management, trading, staking, and token-deployment capabilities through a standardized interface. This allows AI assistants to interact directly with Solana without any custom integration work on the developer's part. The MCP server is also the technical foundation for Burrow, the team's companion no-code product that wraps AgentiPy's capabilities in a Zapier-style automation interface, with an embedded AI assistant called Medusa. Burrow was announced as coming soon in late 2024.

Protocols and Solana Fit

The choice of Solana as the primary target is deliberate. Solana's sub-second finality and sub-cent transaction fees make it one of the few smart-contract platforms where the economics of an AI agent executing many small transactions remain viable. A trading loop that samples prices, rebalances a position, and collects fees multiple times per minute would be cost-prohibitive on Ethereum mainnet; on Solana those same operations are affordable and fast enough to be practically useful.

The protocol coverage reflects the full breadth of Solana's DeFi and NFT stack: Jupiter (swap aggregation), Raydium (AMM and CLMM), Meteora (dynamic liquidity market maker), Orca (concentrated liquidity), Jito (MEV-aware transaction bundling), Drift (perpetuals and spot margin), and Metaplex (NFT standards). DexScreener, Pyth, and Stork handle market data, while Helius provides enhanced RPC and indexing. The framework also supports cross-chain transfers via deBridge, extending the reachable asset universe beyond Solana native tokens.

The APY Token

AgentiPy has a community token, APY, with a fixed supply of one billion. The token functions as a developer-incentive mechanism rather than a protocol-governance or fee-capture instrument. Contributors who build and maintain open-source integrations are eligible for APY grants from the 30% treasury allocation. Ten percent is reserved for the core team. The remaining 60% entered circulating supply starting January 2025 via a gradual release schedule. APY is listed on Gate.io, MEXC, CoinGecko, Dexscreener, and Dextools.

The token model reflects a common pattern in open-source tooling projects: using a fungible reward to sustain contributor momentum in the absence of direct protocol revenue. Whether this is sufficient to sustain long-term protocol coverage depends on actual usage volume and contributor appetite.

Security Posture

AgentiPy recommends environment-variable-based key management rather than hardcoded private keys, and the official documentation explicitly flags the financial risk of autonomous onchain agents. The framework itself does not custody funds; all signing happens locally in the developer's runtime using the keypair they supply. The risk profile tracks that of any autonomous agent with a funded wallet — a buggy or misaligned agent can drain its own funds, particularly in fast-moving DeFi conditions.

Activity and Status

Development has been consistent since launch. PyPI release history shows a December 2024 initial release, a significant 2.0.x overhaul in February 2025, and a sequence of 2.1.x iterations through October 2025. The GitHub repository continues to accept pull requests. The project's X/Twitter account was dormant as of mid-2026 (approximately 262 days since the last post, per third-party uptime monitoring), and the Burrow platform had not been publicly released at the time of writing. The core SDK and MCP server, however, remain functional and installable from PyPI.

AgentiPy occupies a real gap in the Solana developer ecosystem: the intersection of Python-native AI tooling and onchain execution. For developers who want an AI agent to reason over DeFi data and act on it — executing swaps, managing liquidity, monitoring positions, or launching tokens — without the overhead of a Rust or TypeScript integration layer, AgentiPy is one of the few mature, actively maintained options available.