the short version
the Parasol Agent SDK is now MIT-licensed and public on GitHub. the TypeScript types, the example MCP client, the integration guides, the safety documentation — everything a developer needs to build on Parasol's trading intelligence layer.
we did this with grant funding from the Solana Foundation via Superteam UK. since the initial release, we've shipped SDK persistence, 129 tests, a dashboard overhaul, and deep token intelligence — all building on the open-source foundation.
what we open-sourced (and what we didn't)
the SDK is the interface layer. it defines how external AI agents and developers connect to Parasol's intelligence — the tools, the request/response shapes, the authentication model, the MCP protocol integration.
the trading engine behind it — the scoring algorithms, the manipulation filters, the adaptive learner, the risk management system — remains proprietary. that's our product. the SDK is the doorway.
this is the same model used by most production APIs: Stripe publishes client libraries and type definitions, not their fraud detection models. Anthropic publishes the MCP specification, not Claude's weights. the interface is open. the implementation is protected.
why open source the SDK
Superteam UK awarded us $7,500 USDG under their Startup — Early Scaling category. the primary commitment: open source the Parasol Agent SDK on GitHub under MIT license with a working example client and integration guide.
the reasoning is simple. for Parasol's intelligence to become infrastructure that other Solana builders rely on, they need to be able to inspect the interface. they need TypeScript types they can import. they need an example client they can run. they need documentation they can trust because they can read the source.
MIT for the SDK means any developer can build a Parasol integration without legal friction. the intelligence itself is accessed through the hosted API at mcp.parasol.so — same as any other production API.
what's in the public SDK repo
github.com/ParasolSolana/parasol-sdk contains:
example MCP client
a standalone TypeScript client that connects to Parasol's MCP server and demonstrates every tool. clone it, run it, modify it. it has its own package.json and zero dependencies on the main codebase.
``bash
git clone https://github.com/ParasolSolana/parasol-sdk.git
cd parasol-sdk/examples/mcp-client
npm install
npm start
`
10 MCP tools
any AI agent (Claude, GPT, Cursor, custom) can call these via the Model Context Protocol:
| tool | what it does |
|---|---|
parasol_safe_trade_check | pre-trade safety gate: rug risk, honeypot, liquidity, holder analysis — returns SAFE / CAUTION / AVOID |
parasol_analyze_token | full safety analysis with rug risk + honeypot check and clear recommendation |
parasol_check_rug_risk | mint/freeze authority, holder concentration, deployer history |
parasol_check_honeypot | simulated sell test via Jupiter — can you actually exit this token? |
parasol_token_intelligence | deep holder quality analysis: wallet age, coordination detection, funding clusters, deployer rug history. returns a 0-100 score with A-F grade |
parasol_get_trending | scored opportunities from the latest scan cycle with strategy and confidence |
parasol_get_portfolio | current portfolio: open positions, P&L, aggregate performance stats |
parasol_agent_status | agent runtime, cycle count, positions, recent errors |
parasol_start_agent | start autonomous trading |
parasol_stop_agent | emergency stop — open positions remain until manually closed |
the MCP server also exposes 3 resources (parasol://portfolio, parasol://stats, parasol://scanner/latest) and 4 prompts (analyze-token, portfolio-review, risk-assessment, safe-trade) for structured interactions.
13 SDK type definitions
full TypeScript interfaces for every request and response. import them into your project for type-safe integration:
, ApiPermission, RateLimitInfo — authentication and rate limiting — the agent entity with portfolio, stats, and runtime state, UpdateAgentRequest, ExecuteTradeRequest, PushSignalRequest — agent management and signal ingestion, TradeResponse, ScannerResponse, PerformanceResponse — responses with full portfolio, equity curves, and feature importance, SDK_ERROR_CODES — structured error handling with 10 distinct error typesdocumentation
the Phantom integration
one thing we're particularly proud of: Parasol works as a safety layer for Phantom wallet users via MCP.
the flow: an AI agent (Claude, Cursor, any MCP client) calls Parasol's parasol_safe_trade_check before executing a swap through Phantom's MCP server. Parasol checks rug risk, honeypot status, holder quality, and deployer history. if the token is dangerous, the agent stops before the trade reaches the wallet. if it's safe, Phantom previews the swap and the user confirms.
this is what "public good" looks like in practice. Phantom has millions of users. if even a fraction of them avoid one rug pull because an AI agent checked with Parasol first, that's real money saved for real people.
what the grant covered
the full list of deliverables:
all of these are complete and shipped to production on parasol.so.
what we shipped after the grant
the grant deliverables were the foundation. since then, we've continued building on it:
the numbers
| metric | value |
|---|---|
| MCP tools | 10 |
| MCP resources | 3 |
| MCP prompts | 4 |
| REST API endpoints | 8 |
| SDK type definitions | 13 interfaces |
| agent templates | 4 production personalities |
| risk guardrails | 10 independent circuit breakers + concept drift detection |
| documentation pages | 6 developer guides + 20 docs total |
| test coverage | 129 tests, <500ms |
| example client | standalone TypeScript, runs in minutes |
| license | MIT (SDK) |
what this means for builders
if you're building on Solana and need to:
start here: github.com/ParasolSolana/parasol-sdk
if you're using Parasol in your project, we'd love to hear about it. reach out on X @ParasolSolana.