Algorand MCP is a free, open-source server that lets AI assistants work with the Algorand blockchain. It is built for developers and technically curious users who want tools like Claude Desktop, Claude Code, Cursor or Windsurf to read chain data, build transactions and explore Algorand applications without writing every API call by hand. The project is MIT licensed and written in JavaScript and TypeScript.
The server follows the Model Context Protocol, an open standard for connecting AI applications to outside tools and data. Algorand operations are exposed as MCP tools, so any compatible client can call them. The latest release is v4.2.7, and the code was last updated in August 2026. Windows appears in the documentation, but it has not been independently confirmed, so read the release notes before relying on it for anything important.
What Is the Model Context Protocol?
The Model Context Protocol, usually shortened to MCP, is an open standard that defines how AI applications connect to external tools and data sources. Instead of each AI product building its own custom integration for every service, a server exposes capabilities in a common format that any compatible client can use.
In practice, the flow is simple. You ask an assistant a question, the model decides a tool is needed, the client calls the tool through the server, and the server returns structured results. The assistant then uses those results to answer you. Clients such as Claude Desktop, Claude Code, Cursor and Windsurf support this pattern.
For blockchain work, this matters because chain data is precise and easy to misread. A tool that returns exact balances, transaction fields or asset details gives the assistant reliable inputs, which reduces guesswork in its answers.
Who Is Algorand MCP For?
Algorand MCP fits several kinds of users, and each gets something different from it.
- Application developers can ask an assistant to explain an unfamiliar contract, compile TEAL or prototype a transaction before writing final code.
- Data analysts can pull recent transactions or asset activity through the Indexer without learning every query parameter first.
- Students and hobbyists can learn how payments, assets and applications behave on Algorand by watching real results appear in a chat.
- Teams experimenting with agent payments can test x402 paid requests and AP2 workflows in a controlled setup.
For example, a developer building a small marketplace might ask an assistant to check whether a buyer account holds enough of an asset, then build a grouped payment and asset transfer. The developer reviews each step before anything is signed, which keeps the process understandable.
How Agent Wallets Work on Algorand
An agent wallet is an account whose keys are held by software that acts on your behalf. In this server, the mnemonic, a human-readable list of words that recovers a private key, is stored in a local SQLite database. The server uses it to sign transactions when a tool call requires it.
Algorand also supports rekeying, which lets an account delegate signing authority to another address. This is useful for rotating keys or separating everyday operations from a more protected account. Understanding rekeying helps you judge what an agent can actually authorize.
The practical lesson is that a local wallet is only as secure as the computer it runs on. Anyone with access to the database file, or to the operating system account that runs the server, may be able to reach the keys. Treat the database like any other piece of key material, and keep balances in agent wallets modest.
Key features
Agent Wallet and Account Management
Transactions, Atomic Groups and TEAL
Full Algod and Indexer Access
NFDomains Name Resolution
x402 Paid HTTP Requests
AP2 Tooling for Algorand
Tinyman Pools, Swaps and Liquidity
Haystack Router Best-Price Swaps
Supported coins
Coverage centers on Algorand itself, with USDC and ALGO used for x402 payments and for swaps through Tinyman and Haystack Router. The project documentation also mentions Solana (SOL). Confirm each asset in the tool descriptions before you act on it.
How to install Algorand MCP on Windows
- 1
Get the release from the official source
Open the project's GitHub Releases page at github.com/GoPlausible/algorand-mcp and download the latest release, currently v4.2.7. Download only from the official website or repository, not from mirrors or file-sharing sites.
- 2
Verify the file
Compare the SHA-256 checksum of the downloaded file with the value published alongside the release, or verify the signature if one is provided. If no checksum or signature is published, do not install the file until you can confirm its origin.
- 3
Install the program
Unpack or run the release into a folder inside your user profile, such as a folder named algorand-mcp. Keeping it in a location your user account controls makes updates and permissions easier to manage.
-
Connect it to an AI client
Add the server to an MCP-compatible client such as Claude Desktop, Cursor or Windsurf, following that client's server settings. No environment variables are needed for standard use, and network selection happens per tool call.
System requirements
- No environment variables are required for standard use.
- No user action is required for setup, and there are no migration steps.
- Listed platforms: Windows, macOS, Linux and Docker. Windows is mentioned in the documentation but has not been independently verified.
- A compatible MCP client, such as Claude Desktop, Claude Code, Cursor or Windsurf.
How to use Algorand MCP
Using Algorand MCP is mostly about describing what you want in plain language once the server is connected to your AI client. The assistant chooses the right tools, and you review each result before approving any action that involves signing.
- Start with read-only questions, such as asking for an account's balance or the details of a recent transaction, to confirm the connection works.
- Assign a nickname to each account you plan to use so the assistant and you refer to the same wallet.
- Ask the assistant to look up an NFD name if you prefer names over addresses, and confirm the address it returns.
- Request a transaction build before signing, and read the sender, receiver, amount and asset fields.
- For swaps, ask for a quote from the Haystack Router and compare it with the Tinyman pool price.
- Review the final transaction summary, then approve it only if every field matches your intent.
Pros & cons
- Free and MIT licensed, with public source code on GitHub.
- Covers core Algorand operations, including payments, assets, applications and key management.
- Works with popular AI clients such as Claude Desktop, Cursor and Windsurf.
- Includes Tinyman and Haystack Router integrations for swaps and liquidity.
- The latest code update is dated August 2026, showing ongoing development.
- Windows support is mentioned in the documentation but has not been independently verified.
- The agent wallet keeps mnemonics in a local database, so the security of your computer matters a great deal.
- Requires comfort with MCP clients and a basic understanding of Algorand accounts.
- Results depend on the assistant interpreting your request correctly, so review each transaction carefully.
Is Algorand MCP safe?
Algorand MCP handles wallet keys, so treat installation as a security decision. Download it only from the official GitHub repository at github.com/GoPlausible/algorand-mcp or its Releases page. Compare the SHA-256 checksum, or verify the signature, against what the release publishes before running anything.
Never type a seed phrase or mnemonic into a third-party website, a chat tool or an AI prompt. The server stores mnemonics in a local SQLite database and does not return them in tool responses, but that protection depends on your machine staying clean. Use a dedicated test account first, keep agent balances modest, and check the latest version on the official website before updating.
- Official source only. Download from github.com/GoPlausible/algorand-mcp. Ignore ads and lookalike domains.
- Check the SHA-256. Compare the installer hash with the value the developer publishes before running it.
- Never share your seed phrase. No developer, support agent or website — including CryptApps — will ever ask for it.
Alternatives
Algorand MCP is not a node and not a trading terminal. It is a bridge that lets AI clients call Algorand operations. GO Algorand is the node software itself, which makes it the better choice if you need to run infrastructure. Algorand Node UI is a terminal interface for managing remote nodes, aimed at operators rather than AI-driven workflows.
Surfpool and Agave belong to the Solana ecosystem, so they serve a different chain. Surfpool is a local development environment, and Agave is a validator. If you want an assistant to read Algorand data or build transactions for you, Algorand MCP is the tool designed for that job.
Specifications
- Price
- Free
- Type
- Open source
- License
- MIT
- Latest version
- v4.2.7
- Last update
- 8 Aug 2026
- Windows support
- Unverified (Windows mentioned in docs)
- Source code
- github.com/GoPlausible/algorand-mcp
Frequently asked questions
Yes. Algorand MCP is free and released under the MIT license, which allows use, modification and redistribution under its terms. The source code is public on GitHub, so you can review what the server does before running it. Paid x402 requests spend USDC or ALGO from your wallet, which is separate from the cost of the software itself.
Algorand MCP is designed for clients that support the Model Context Protocol, including Claude Desktop, Claude Code, Cursor and Windsurf. Other MCP-compatible clients should also work. Each client registers servers in its own way, so follow that client's settings for adding a new server.
It can be, if you handle it carefully. Download it from the official GitHub repository, verify the checksum or signature, and never enter a seed phrase into a third-party site. The agent wallet keeps mnemonics in a local database, so a compromised computer is a serious risk. Start with a test account before connecting any real funds.
Windows is listed in the project documentation alongside macOS, Linux and Docker. We have not independently verified Windows support, so test with a small, low-stakes setup first. Check the GitHub Releases page for notes about Windows builds before relying on the server for daily work.
It stores mnemonics for agent wallets in a local SQLite database so it can sign transactions when a tool call requires it. The server is designed not to return mnemonics in tool responses. The phrase still exists on disk, so protect the machine and the database file like any other key material.
Yes, in practice. Algorand MCP reads network data through the Algod and Indexer APIs, so it needs access to a node or an endpoint that provides those APIs. You can use a node you already run, or a service that exposes the same interfaces. The server itself does not replace the node software.
Verdict
Algorand MCP suits developers and advanced users who want an AI assistant to read Algorand data, build transactions and explore integrations such as Tinyman and Haystack Router. It is free, MIT licensed and actively updated, with the latest code change dated August 2026. The agent wallet adds convenience, but it also means keys live on your computer, so start with test accounts and keep the software current. If you need to run a node rather than an AI bridge, GO Algorand is the better fit.