The Vaults.fyi MCP server connects MCP-compatible clients to the full Vaults.fyi data and transaction surface. Discovery, history, recommendations, position tracking, and transaction building are all callable as tools — your client decides which to use and how to chain them. Hosted atDocumentation Index
Fetch the complete documentation index at: https://docs.vaults.fyi/llms.txt
Use this file to discover all available pages before exploring further.
https://mcp.vaults.fyi/mcp over streamable HTTP. 38 tools. Bearer-token auth. No server-side private keys: every transaction tool returns calldata for your client to sign.
What people use it for
A few prompts that exercise meaningful tool chains:Discover yield
Discover yield
“Find me the best USDC yields right now with TVL over $10M and no active warnings.”
“What yields beat the USDC benchmark by more than 200 bps over the last 30 days?”
Filter by curator or protocol
Filter by curator or protocol
“Show me Steakhouse and Gauntlet curated Morpho vaults across all stablecoins.”
“Which Aave v3 markets have the highest USDC supply APY across networks?”
Model layered yield
Model layered yield
“I’d consider sUSDe and sUSDS — find vaults with composite APY above 8% accounting for the underlying asset rate.”
Run history and benchmarks
Run history and benchmarks
“Plot Fluid USD Vault APY against the USDC benchmark for the last 90 days.”
Audit positions
Audit positions
“Audit positions for 0xabc… — total deposited, current value, unclaimed rewards, and any flagged exposure.”
Build and execute transactions
Build and execute transactions
“Build a deposit of 10,000 USDC into the highest-rated Morpho vault on Base and return the calldata.”
Quickstart
You’ll need a Vaults.fyi API key. Request one at portal.vaults.fyi.
- Claude Desktop
- Claude Code
- Cursor & other clients
Add the server to your Claude Desktop config (Restart Claude Desktop. The 38 Vaults.fyi tools will appear in the tools menu.
~/Library/Application Support/Claude/claude_desktop_config.json on macOS, %APPDATA%\Claude\claude_desktop_config.json on Windows):claude_desktop_config.json
What you can do
Six capability surfaces, each backed by several tools:Discover and filter
Search across networks, protocols, assets, curators, and tags. Filter by TVL, APY, score, and active warnings in a single structured query.
Evaluate risk
Reputation Scores and inline flags (bad debt exposure, qualified-investor gates, depegs, recent incidents) ship with the data. No separate lookup.
Analyze history
APY, TVL, share-price, and total-return time series for any vault. Asset-price history and benchmark APY for relative performance.
Build and execute transactions
Deposit, withdraw, claim rewards, and approve. Tools return decoded calldata and signing instructions; your client signs and submits.
Track portfolios
Aggregate positions and rewards across networks for any address. Drill into per-position cost basis and event history.
Operate on tokens
Approve, transfer, wrap, and unwrap as primitives — useful when an agent is composing multi-step flows.
Tool inventory
38 tools, grouped by purpose. Each returns structured JSON optimized for agent consumption.Discovery and search (7)
Discovery and search (7)
| Tool | Purpose |
|---|---|
vaults_search | Filtered search across vaults (TVL, APY, score, curator, tags, warnings) |
vaults_list | Paginated full vault list |
networks | Supported chains |
protocols | Supported protocols |
assets | Supported assets |
curators | Named curators (79+) |
tags | Vault tags (e.g., top-curator, lst, rwa) |
Vault detail (3)
Vault detail (3)
| Tool | Purpose |
|---|---|
vault_details | Full vault metadata, including flags and Reputation Score |
vault_apy | Current APY (headline, base, rewards, composite) |
vault_tvl | Current TVL |
History and benchmarks (8)
History and benchmarks (8)
| Tool | Purpose |
|---|---|
vault_apy_history | APY time series |
vault_tvl_history | TVL time series |
vault_share_price_history | Share-price time series |
vault_history | Combined APY/TVL/share-price |
vault_returns | Realized total returns |
asset_price_history | Underlying asset price |
benchmark_apy | Asset-level benchmark APY |
benchmark_apy_history | Benchmark APY time series |
Recommendations (2)
Recommendations (2)
| Tool | Purpose |
|---|---|
best_vault | Single best-fit vault for a given asset/criteria |
best_deposit_options | Ranked deposit options |
Wallet and positions (6)
Wallet and positions (6)
| Tool | Purpose |
|---|---|
positions | All positions for an address |
position_details | Per-position cost basis, current value, P&L |
wallet_balances | Token balances across networks |
token_balance | Single token balance |
user_events | Deposit/withdraw/claim history |
rewards | Outstanding reward entitlements |
Vault transactions (7)
Vault transactions (7)
| Tool | Purpose |
|---|---|
transaction_context | Pre-flight context (allowance, balances, gas) |
build_vault_tx | Deposit/withdraw calldata with decoded steps |
build_claim_rewards | Claim calldata for a single vault |
claim_all_rewards | Multi-vault claim calldata |
withdraw_all_positions | Multi-vault withdraw calldata |
submit_tx_hash | Register a submitted hash for status tracking |
get_transaction_status | Status of a tracked transaction |
Token utilities (5)
Token utilities (5)
| Tool | Purpose |
|---|---|
approve_erc20 | ERC-20 approval calldata |
transfer_erc20 | ERC-20 transfer calldata |
transfer_native | Native transfer calldata |
wrap_native | Wrap native to wrapped equivalent |
unwrap_native | Unwrap |
Authentication
Every request requires a bearer token:The API key is forwarded only to Vaults.fyi API calls and is never placed in signing URLs or transaction payloads.
Signing model
Transaction tools (build_vault_tx, build_claim_rewards, approve_erc20, etc.) return structured calldata, not signed transactions. The server holds no private keys.
Each response includes:
stepDecodings— human-readable description of each callsigningInstructions— step-by-step guidance for the signersigningPaths— supported signing surfaces (e.g., wallet, OWS)submitTool— the tool to call after signing (submit_tx_hash)statusTool— the tool to call for status (get_transaction_status)
Going further
AI agents overview
The full surface area for agents: MCP, x402, and the LLM-optimized API reference.
Keyless access via x402
Pay-per-request access on Base, no API key required.
LLM-optimized API reference
Full reference at
llms.txt and llms-full.txt.OpenAPI spec
Underlying REST API the MCP wraps.

