How it works
Explore how an AI client can work with Solana data through a tool interface. Read accounts, balances and transactions, understand instructions, and review simulation results. This page explains the request flow and integration design. The website provides reference material for configuring a separately deployed MCP service.
MCP, the Model Context Protocol, is an open protocol for connecting AI clients to tools. A client that supports it, such as Claude, Claude Code, ChatGPT, Cursor, VS Code or Devin Desktop, can be pointed at an MCP server and will then see that server's tools, with a description of what each one does and what it needs. When you ask for something, the AI decides which tool fits, calls it, and uses the answer in its reply. The examples here describe a Solana tool interface. A service advertises its own tools and limits, and a client must discover that actual tool list after connecting.
The path of a request
What happens after you ask a question in your AI client.
You type a request in your AI client in plain words, for example “why did my last deposit fail?”.
The client discovers the service's advertised schemas. It can select solana_get_transaction with a signature, then inspect available instructions and logs to explain a failure.
A configured service validates the request, calls the selected Solana RPC and returns structured data with its network and commitment context.
The AI uses the returned records in its answer. Logs, errors and missing data should stay visible so the explanation can be checked against the transaction.
Twelve groups of tools
Reference schemas and designs, with protocol context.
Read and simulate
Nine reference schemas describe Solana account, token, transaction and network data, plus transaction simulation. They explain RPC inputs and outputs for a separately configured service.
solana_get_balance
solana_get_token_balances
solana_get_account_info
solana_get_signatures
solana_get_transaction
solana_get_signature_status
solana_get_token_supply
solana_network_status
solana_simulate_transaction
Understand
Read the structure of a transaction and compare its intended effects with the available chain data. Simulation, decoding and log inspection answer different questions and carry different limits.
solana_simulate_transaction
solana_decode_transaction
solana_trace_call
Build
A development workflow starts with reviewed Rust or Anchor source, reports compiler diagnostics, estimates transaction costs and verifies deployed binaries. These schemas outline a developer integration.
solana_scaffold
solana_program_build
solana_estimate_fees
solana_deploy
solana_verify
Act
Transaction preparation, signing, submission and confirmation are separate steps. This design reference makes the required authorities, network and expected account changes explicit at each step.
solana_prepare_transaction
solana_prepare_batch
solana_send_signed
solana_wait
solana_key_status
solana_agent_wallet
Agents
Agent integrations need explicit identity, service discovery, payment authorization and event processing. The examples describe application designs and avoid assuming a single Solana agent registry or reputation standard.
solana_agent_register
solana_agent_lookup
solana_agent_feedback
solana_agent_pay
solana_agent_job
solana_watch
solana_react
solana_index
Drift perps
Drift perpetuals use Solana accounts, oracle prices, margin and funding. Liquidity can come from its decentralized orderbook, just-in-time makers and AMM. The reference separates market data from order execution.
solana_perps_markets
solana_perps_market
solana_perps_account
solana_perps_funding
solana_perps_candles
solana_perps_risk
solana_perps_plan
solana_perps_scan
solana_perps_backtest
solana_perps_agent
solana_perps_submit
Privacy and data
Solana account activity is publicly observable. Privacy designs must specify the additional protocol, threat model and metadata exposure. Local preparation alone does not conceal an on-chain sender, recipient or amount.
solana_private_pool
solana_private_deposit
solana_private_withdraw
solana_private_invoice
solana_private_risk
solana_private_disclose
solana_private_session
Jupiter Swap
Jupiter aggregates token-swap liquidity on Solana. Quotes describe routes, amounts, fees and execution constraints, while transaction instructions support a separate wallet-signing flow. Liquidity positions belong to the underlying DEX.
solana_jupiter_tokens
solana_jupiter_quote
solana_jupiter_swap
solana_jupiter_swap_instructions
solana_jupiter_route
solana_jupiter_price
solana_jupiter_order_status
solana_jupiter_transaction_review
Raydium LaunchLab
LaunchLab is Raydium's token-launch product. Tokens trade along configured bonding curves before eligible launches graduate into Raydium liquidity pools. Launch parameters, fees and migration rules belong to that protocol.
solana_raydium_pools
solana_raydium_quote
solana_raydium_swap
solana_raydium_launchlab
Wormhole
Wormhole supports cross-chain messaging and token-transfer integrations, including Solana. A transfer spans source and destination chains with separate finality, verification and redemption steps. Routing support depends on the chosen product.
Kamino
Kamino provides Solana lending markets and vault strategies. Review reserve utilization, collateral risk, borrow positions and strategy parameters. Yields vary with market activity, fees and the selected product.
solana_kamino_markets
solana_kamino_obligation
solana_kamino_vaults
solana_kamino_deposit
solana_kamino_withdraw
Pyth price data
Oracle prices need a feed identifier, publish time, exponent and confidence interval. Pyth provides price-update products for Solana, each with its own verification, freshness and access requirements.
The catalogue separates RPC references, application designs and protocol mechanics. Tool names are illustrative schemas. A connected service determines which calls are actually available.
Explore tool references →Review a transaction
Understand the message, signers and simulation before using a separate wallet application.
Inspect the actual message
Decode the transaction bytes and inspect the fee payer, signers, program IDs and token changes. A readable preview should be tied to that message before any wallet authorizes it.
Define delegated authority
A smart-wallet design can define spending caps, program scope and expiry. These limits depend on the selected program and its enforcement. Review revocation and fee funding as part of the authority model.
Compare state and outcome
Simulation is an observation at a point in time. Recheck the network, blockhash and relevant account state before submission, then inspect confirmation and execution errors afterward.
Connect with clear context
Read, understand and simulate on the intended cluster.
Solana
Accounts, balances, instructions and simulation form the core Solana reference. Select the cluster and commitment explicitly.
A configured service
An MCP client needs a service URL and transport settings. Its discovered tool list determines the available operations and network scope.
Discover the schema
Check each tool's input schema, permissions, error format and data source. Similar names can still have different service limits.
Pyth price data
Review feed IDs, publication times and confidence intervals. Oracle freshness and verification depend on the selected Pyth product.