AI coding agents can already write smart contracts and explain block explorers. The harder step is letting them act on a live network without handing over a private key. The Hedera Network MCP attacks that problem directly. It gives agents a single endpoint for building Hedera transactions, while the user’s own wallet keeps final authority. As a result, an agent can propose a transfer or a token mint, but it cannot complete either one alone.
The Hedera Network MCP is live on testnet. Connect Claude Code, Codex, Cursor, ChatGPT & more with one endpoint, then ask your agent to send HBAR, mint tokens, or query the network. Nothing settles until you sign it in your own wallet. Try the Hedera Network MCP: https://t.co/T9J8O1QtZG
— Hedera (@hedera) October 21, 2026
One Endpoint for Every Major Coding Agent
The Hedera Network MCP runs on Hedera Testnet at https://agentic-testnet-mcp.hedera.com/mcp. It uses the Model Context Protocol, the open standard that lets AI clients discover and call external tools. Hedera lists Claude Code, Codex, Cursor, VS Code, Devin, Claude on web and desktop, and ChatGPT as supported clients. Setup stays short in most of them. Claude Code needs one command, claude mcp add --transport http, and Codex CLI uses a similar codex mcp add line. Cursor and VS Code read the endpoint from a JSON config file. Notably, ChatGPT Desktop only works with the server from its Work tab, so users should expect small differences between clients.
This design removes a familiar integration burden. Previously, a team that wanted agents on Hedera often wrote custom glue code for each AI framework. Now the same endpoint serves every compatible client. Developers add one URL and authenticate once. In contrast to a local setup, they run no agent infrastructure of their own. The server is a managed, hosted instance of the Hedera Agent Kit, the toolkit Hedera built for agent integrations.
What the Agent Can Actually Do
Hedera’s server page describes more than 40 tools in four groups. The account group covers HBAR transfers, account creation, and allowance management. The token group lets agents create, manage, and airdrop tokens through the Hedera Token Service. Additionally, the docs list minting, association, and ERC-20 and ERC-721 contract queries. The consensus group handles topics and messages on the Hedera Consensus Service. Finally, the query group reads balances, accounts, and topic data from mirror nodes.
A developer could therefore ask an agent to create a test token, mint a supply, and airdrop it to a second account. The agent would chain those calls and explain each step in plain language. Meanwhile, read-only questions need no approval at all. Asking for an account balance or the current exchange rate returns data immediately. Only state-changing actions trigger the signing flow. That split keeps low-risk exploration fast and high-risk actions deliberate.
How Signing Works, and How It Changed
The signing model carries the most important design choices. Hedera’s July 7 launch post described a server in RETURN_BYTES mode. The server built a transaction, froze it, and returned hex-encoded bytes. The developer then signed and submitted those bytes with a script or the Hedera SDK. The post summed up the principle in one line: the agent “can propose, but it cannot spend, until you sign.”
The current documentation describes a more integrated flow. Users now authenticate through OAuth 2.1 with Dynamic Client Registration, so clients need no pre-registered ID. The first tool call returns a 401 error, which starts a WalletConnect session. The user scans a QR code or uses a browser extension, then signs a single-use sign-in message. No funds move at that step. The server checks the signature against the account’s key on the mirror node, then issues a token scoped to that account.
After sign-in, transaction tools build and freeze the transaction and send it to the user’s wallet. The tool returns a requestId, and the agent polls a check_signing_status tool. Statuses include awaiting_approval, executed, rejected, and expired, with expiry after about 180 seconds. Importantly, responses carry a walletVerifiable flag. Contract calls and file operations return false, so the agent should show the user a plain summary. Users need a funded testnet account and a wallet such as HashPack or Kabila.
Limits Builders Should Know
Hedera does not oversell the current release. The server works on Testnet only, so teams that want mainnet access must self-host the Agent Kit MCP server. The docs also state that the server offers no spend policies, caps, or recipient allowlists. The wallet approval prompt remains the only spending control. Furthermore, the server supports single-key accounts only, so threshold and key-list accounts cannot connect. Sessions also expire without a proactive prompt, which forces users to disconnect and reconnect.
Those limits point to how Hedera plans to grow the product line. The same page lists a Hedera Payments MCP as coming soon, designed to let an assistant pay within limits the user sets. That product would address the missing spend controls. Hedera also offers a Docs MCP at https://docs.hedera.com/mcp, which gives agents read-only access to live documentation with no wallet. In addition, an open-source Hedera Skills marketplace packages instructions and references for coding agents. Together, these pieces cover network access, documentation, and development guidance.
Why the Design Matters
The wider industry still debates how much autonomy agents should hold onchain. Some projects give agents their own keys and accept the risk. Hedera chose the opposite default for its hosted server. The agent handles the tedious work of constructing a correct transaction, and the human handles authorization. That arrangement suits testnet experimentation well, because developers can iterate quickly without exposing real assets.
The harder test comes next. As the launch post itself noted, safe onchain action still depends on human-side signing and integration discipline. Builders will decide whether wallet prompts stay readable when agents generate dozens of requests. They will also decide whether spend limits arrive before mainnet demand does. For now, the Hedera Network MCP gives developers a low-friction way to test those questions on a network built for low fees and fast finality.
*Disclaimer: News content provided by Genfinity is intended solely for informational purposes. While we strive to deliver accurate and up-to-date information, we do not offer financial or legal advice of any kind. Readers are encouraged to conduct their own research and consult with qualified professionals before making any financial or legal decisions. Genfinity disclaims any responsibility for actions taken based on the information presented in our articles. Our commitment is to share knowledge, foster discussion, and contribute to a better understanding of the topics covered in our articles. We advise our readers to exercise caution and diligence when seeking information or making decisions based on the content we provide.
