Search
⌘K
Overview
SDKs
Available SDKs
AI Integration
Ecosystem
Supported Clients
APIs
Authentication
Swap API
Cross-Chain Swaps (Fusion+)
Intent Swap (Fusion)
Classic Swap
Swap RWAs
Orderbook API
History API
Traces API
Portfolio API
Balance API
Gas Price API
Spot Price API
Token API
NFT API
Transaction Gateway API
Charts API
Domains API
Token Details API
Web3 RPC API
Introduction
Arbitrum
Avalanche
Base
BNB Chain
Cronos
Ethereum
Gnosis
Linea
Monad
Optimism
Polygon
Solana
Sonic
Unichain
ZKsync
Resolvers
Introduction
Terms of Use Resolver Terms
Limit Order
Quick Start
Contract interaction
Resolver farming guide
Unicorn Power setup
Intent/Fusion
Evm quicklinks
Solana quicklinks
Cross Chain
Cross chain examples
Exclusive Resolver API Terms
FAQ
API Error Messages
API Troubleshooting
Cors errors
Tx input data
Infrastructure fees
Docs·AI Integration·Ecosystem

Ecosystem

1inch provides two complementary AI integration methods: the MCP Server — a live protocol your AI assistant uses to execute swaps, place limit orders, query APIs, and access documentation — and AI Skills — reusable instruction packages that teach agents the setup patterns. Both are free to connect, work together, and share the same authentication.

MCP Server

The 1inch MCP Server is an AI-native integration: your editor or CLI agent speaks to 1inch over the Model Context Protocol (MCP) so it can execute token swaps, place limit orders, query portfolio and price data, and pull documentation — all through natural conversation.

Cost and access: There is no separate charge to connect to the MCP server. Swap, Orderbook, Business API access, and the debug tool use the same authentication and API entitlements as the rest of the 1inch Business / developer stack (standard API terms apply to those calls). Documentation search, the example catalog, and example source work without authentication, so you can start exploring immediately.

The server is hosted at https://api.1inch.com/mcp/protocol

For setup instructions per client, see Supported Clients.

For AI agents and automation that don't use MCP, see our LLM-optimized documentation -- a machine-readable index of all APIs following the llmstxt.org specification.

MCP Server capabilities

The MCP server exposes seven tools. The first three are public; the rest require authentication.

Tool Auth Description
swap Required Build quotes and execution steps for classic (Pro), intent (Fusion), and cross-chain swaps
orderbook Required Build, sign, submit, list, and cancel limit orders via the 1inch Orderbook API
product_api Required Call 1inch Business product APIs (Portfolio, Spot Price, Token, Gas Price, Web3 RPC, and more)
debug Required Debug production issues: look up request logs by correlation id or time window
search Public Search documentation and API reference
list_examples Public List available SDK code examples
get_example Public Retrieve full example source (TypeScript, Go, Python, etc.)

Full parameter reference: Available tools.

AI Skills

The 1inch AI Skill is a reusable instruction package that teaches AI agents how to connect to and use the 1inch MCP server — the production URL, available tools, authentication setup, and client configuration patterns. Install it once and your agent knows the setup automatically.

Bash
1
npx skills add 1inch/1inch-ai

Source and full details: github.com/1inch/1inch-ai

Use both together: the skill teaches your agent how to connect; MCP provides the live connection.

Authentication

The public tools (search, list_examples, get_example) work without authentication. The execution tools (swap, orderbook, product_api, debug) require authentication via either an API key or OAuth.

API key

Pass your 1inch API key via the Authorization: Bearer <apikey> header. Get a key from the 1inch Business Portal.

How to set the header depends on your MCP client — see the Authentication setup section in Supported Clients for client-specific configuration.

OAuth

When you call an authenticated tool without an API key, MCP clients that support OAuth will automatically prompt you to log in via the 1inch Business Portal. After authentication, all tools become available for the duration of your session. For all supported OAuth flows including client_credentials for server-to-server use, see OAuth 2.1.

Agentic use

"Agentic" here means your client runs a tool loop: the model decides which MCP tools to call, uses the results, and may call again until the task is done. You describe the outcome; the agent plans steps.

Typical tool chains

Goal How the agent usually proceeds
Swap or quote exploration Swap tool for quotes / steps; combine with doc search for mode differences (classic vs Fusion vs cross-chain)
Place a limit order Orderbook tool to build the order server-side → sign the returned typed data → submit back via orderbook
Operational checks Business API tool for token metadata, gas, spot price, portfolio P&L, etc., then map results back into your code
Debug production issues debug with a request id from API responses or a time window to retrieve logs for your account
New feature in your repo Documentation search → example catalog → example source → apply or refactor code in your project
Correct API usage Search docs for parameters and errors → cross-check with an example → suggest headers, chains, and payloads

You do not need to name each tool: a capable client exposes them to the model; prompts like "swap 100 USDC to ETH on Base" or "find the official swap example" are enough.

Prompt patterns that work well for agents

  • Direct execution: "Swap 100 USDC to ETH on Base from my wallet 0x1234…"
  • Multi-step flows: "Place a limit order to sell 1000 USDC for 1000 USDT on Ethereum, expiring in 1 hour."
  • Outcome first: "Implement ERC-20 → USDC swap on Base in TypeScript; use 1inch MCP for docs and examples, then wire it into src/swap.ts."
  • Operational queries: "What's my portfolio P&L on Ethereum?" or "What's the current gas price on Base?"
  • Verify against examples: "Fetch the closest SDK example from MCP, then diff our quote() call against it."

MCP vs llms.txt for agents

MCP /llms.txt
Best for Agentic execution (swaps, orders, APIs) + interactive coding assistants Bulk context, custom RAG, CI jobs, clients without MCP
Freshness Live API access + server-side search (periodically indexed) Snapshot index you fetch and chunk yourself
Authenticated APIs Yes — swaps, orderbook, portfolio, gas, spot price, debugging, and more No — static links and summaries only

Use both if you split work: e.g. llms.txt in a pipeline for broad grounding, MCP in the IDE for grounded examples and tool-backed answers.

Limits and responsibility

  • Sessions: See Too many sessions (429)—each client window may count as a session; agents should avoid spawning unbounded parallel MCP connections.
  • Approvals: Your client may ask you to approve tool calls, especially OAuth or API writes. The MCP server does not bypass your org's access rules.
  • Secrets: Pass API keys or headers only as documented for your client (e.g. supergateway --header); never commit credentials.

Available tools

The 1inch MCP Server provides seven tools. Four execution tools require authentication; three public tools work without it.

Authenticated tools (API key or OAuth)

These tools require authentication. You can authenticate via an API key header or OAuth — see Authentication setup for client-specific configuration.

swap

Execute token swaps via 1inch. The server automatically finds the best rate across all available routes, handles token resolution, allowance checks, and transaction building — the agent only signs transactions where required. Supports same-chain and cross-chain swaps.

Parameter Type Required Description
src string Yes Source token (symbol or address)
dst string Yes Destination token (symbol or address)
amount string Yes Amount in smallest units
chain number Yes Chain ID
from string Yes Sender wallet address
quoteOnly boolean No If true, returns quotes without building a transaction
preferredType string No "classic", "fusion", or "crosschain" (omit to auto-route)
dstChain number No Destination chain ID (triggers cross-chain mode)
slippage number No Slippage tolerance in percent (default 0.5)
signedOrder string No Submit a previously signed Fusion/cross-chain order
orderHash string No Order hash for submission

Typical flow:

  1. Call swap with source/destination tokens and amount — the server finds the best rate automatically
  2. Sign the returned transaction (eth_sendTransaction) or typed data (eth_signTypedData_v4)
  3. For Fusion/cross-chain: call swap again with the signedOrder to submit

orderbook

Manage limit orders via the 1inch Orderbook API (v4.1). The server builds orders server-side using the Limit Order SDK — no SDK installation needed on the client.

Actions:

Action Description
build Server-side order construction. Returns EIP-712 typed data for signing
create Submit a signed order to the orderbook
list Query orders by maker address, order hash, or browse all with pagination
cancel Load order state and return on-chain cancellation guidance

Build parameters:

Parameter Type Required Description
action string Yes "build"
chain number Yes Chain ID
makerAsset string Yes Token to sell (symbol or address)
takerAsset string Yes Token to receive (symbol or address)
makingAmount string Yes Amount to sell (smallest units)
takingAmount string Yes Amount to receive (smallest units)
maker string Yes Maker wallet address
expirationSeconds number No Expiration in seconds (default 600)

Typical flow:

  1. Call orderbook with action: "build" and your order parameters
  2. Approve the maker asset for the Limit Order Protocol contract (address provided in the response)
  3. Sign the returned typedData with eth_signTypedData_v4
  4. Call orderbook with action: "create", passing orderHash, signature, and orderData from the build response

product_api

Call any 1inch Business product API endpoint using your authenticated session. This single tool provides access to all 1inch APIs: Portfolio, Spot Price, Token, Gas Price, Web3 RPC, Balance, NFT, Charts, Transaction Gateway, and every other endpoint available in the 1inch Business platform.

For Portfolio v5.0 through this gateway, use paths starting with /portfolio/portfolio/v5.0/... (the first /portfolio segment is stripped when routing to the Portfolio service).

Parameter Type Required Description
method string No "GET" or "POST" (default "GET")
path string Yes API path (e.g. /portfolio/portfolio/v5.0/general/current_value)
query object No Query parameters as key-value pairs
body object No Request body for POST requests

debug

Debug production issues by looking up request logs. Provide a correlation id (x-request-id from any API response) to trace a specific request, or specify a time window to retrieve all logs for your account.

Parameter Type Required Description
requestId string No* x-request-id (or correlation id) from API responses. When set, searches logs for this value; optional startTime / endTime narrow the window (defaults to the last 24 hours ending at endTime or now).
startTime string No* RFC3339 start of the query window (required with endTime when requestId is omitted).
endTime string No RFC3339 end of the query window.
logLevel string No Filter by log level: "info", "warn", or "error". Omit to return all levels. Only applies to time-window mode (ignored when requestId is set).
limit number No Max log lines (1–100, default 50).

*Provide requestId, or both startTime and endTime, to list all logs in that window (when requestId is omitted).

Log retention: The time window you can query depends on the log retention period of your subscription plan.

Response shape: The tool returns JSON with mode (request_id or logs), timeRange (start / end in RFC3339), count, and logs (array of entries). Each entry includes timestamp, message, and optional level and HTTP status code.

Public tools (no authentication required)

Search across all 1inch documentation, API references, and SDK guides. Returns matching results with titles, snippets, and links.

Parameter Type Required Description
query string Yes Search query
limit number No Results per page (1–100)
page number No Page number (1-based)
include_body boolean No Include full document body in results

list_examples

List all available SDK code examples. Returns example names that can be used with the get_example tool. No parameters required.

Currently available examples:

  • @1inch-examples/intent-swap — Intent/Fusion swaps for EVM and Solana
  • @1inch-examples/limit-order — Limit orders for EVM chains
  • @1inch-examples/cross-chain-swap — Cross-chain swaps (Fusion+)

get_example

Retrieve the full source code of an SDK example.

Parameter Type Required Description
name string Yes Example name from list_examples
file string No Specific file path (omit to get all files)

What you can do

Once connected, you work through natural conversation. Your AI assistant calls the right tools behind the scenes.

Execute token swaps

Describe the swap you want ("swap 100 USDC to ETH on Base"), and the assistant automatically finds the best rate across all available routes, handles token resolution, allowance checks, and transaction building. You only need to sign the transaction. Supports same-chain swaps (classic and Fusion) and cross-chain swaps.

Place limit orders

Create limit orders without installing any SDK. Tell your AI "place a limit order to sell 100 USDC for 100 USDT on Ethereum", and the server builds the order server-side, returns the data for you to sign, and submits it to the orderbook.

Query any 1inch API

The product_api tool gives your AI access to all 1inch Business product APIs — Portfolio, Spot Price, Token, Gas Price, Web3 RPC, Balance, NFT, Charts, Transaction Gateway, and more. Ask "what's my portfolio value on Ethereum?" or "what's the current gas price on Base?" — every API endpoint available in the 1inch Business platform is accessible through this single tool.

Debug production issues

Use debug to trace a request by x-request-id or to list logs in a time range. Paste the correlation id from a failed API response and ask the assistant to pull the matching log lines.

Search documentation and browse SDK examples

Ask any question about 1inch APIs, supported chains, error codes, or integration patterns. Your AI searches across all documentation, API references, and SDK guides — no manual browsing required. You can also pull complete, working code examples directly into your conversation in TypeScript, Go, or Python.

The examples are provided in TypeScript — but your AI assistant can translate them into any language. Just ask for "the swap example in Java" or "convert this to Rust" and it will adapt the code for you.

Example prompts

The following examples show the kinds of questions and requests that work well with the 1inch MCP Server and AI Skills. Use them as starting points and adapt to your specific use case.

Executing swaps

"Swap 1000 USDC to ETH on Ethereum from my wallet 0x1234..."

"Find the best rate for swapping 0.5 ETH to USDT on Arbitrum."

"Swap 100 USDC from Base to Polygon."

The assistant calls the swap tool, automatically finds the best rate and route, and returns the transaction or typed data for you to sign.

Managing limit orders

"Place a limit order to sell 1000 USDC for 1000 USDT on Ethereum, expiring in 1 hour."

"List my active limit orders on Ethereum for wallet 0x1234..."

"Check the status of order 0xabc..."

The assistant builds orders server-side, guides you through signing, and submits to the orderbook.

Querying portfolio and token data

"What's my portfolio P&L on Ethereum for wallet 0x1234...?"

"What's the current price of 1INCH in USD?"

"What's the gas price on Base right now?"

The assistant calls the appropriate product API endpoint and returns the data directly.

Debugging with production logs

"Here's my x-request-id from a 500 response — pull the server logs."

"List logs between 2026-01-15T00:00:00Z and 2026-01-15T23:59:59Z."

The assistant calls debug with your request id or time window and returns the matching log entries.

Debugging and troubleshooting

"My swap transaction is returning a 400 error. What are the common causes?"

"I'm getting CORS errors when calling the 1inch API from my frontend. How do I fix this?"

"The approval transaction succeeded but the swap still fails with 'insufficient allowance'. What should I check?"

The assistant searches the FAQ and API documentation for known issues and solutions specific to your error.

Starting a new integration

"I'm building a swap feature on Base. Show me a complete TypeScript example for swapping ERC-20 tokens using the 1inch API."

"I need to integrate 1inch swaps into my Go backend service. Show me how to use the Go SDK."

"How do I call the 1inch swap API from Python?"

The assistant retrieves the relevant SDK example with full source code — including wallet setup, token approval, swap execution, and error handling — in the language you need, ready for you to adapt.

Getting code for specific scenarios

"Show me how to swap native ETH to USDC on Base, including gas estimation."

"Get the Solana swap implementation with proper error handling."

"How do I check and set token allowance before executing a swap in Go?"

"Show me how to set up a viem provider configured for 1inch on Arbitrum."

"Show me a working cross-chain swap example between Ethereum and Base."

The assistant retrieves the matching documentation and example files in your language, and highlights the relevant sections for your scenario.

Understanding APIs and parameters

"What authentication method does the 1inch API use? Show me the header format."

"What parameters does the swap endpoint accept, and which ones are required?"

"What chains and chain IDs does 1inch support?"

"How does the orderbook API work for limit orders?"

The assistant searches the API reference and returns the specific details you need, with links to the full documentation for deeper reading.

Prompting tips

To get the most out of the MCP Server:

  • Just say what you want done. "Swap 100 USDC to ETH on Base" or "place a limit order for 1000 USDT on Ethereum" — the assistant handles the rest.
  • Mention your chain. Saying "on Base", "on Arbitrum", or "cross-chain from Ethereum to Polygon" helps the assistant pick the right route immediately.
  • Combine execution with research. The assistant maintains context, so you can ask "what's the best rate for this swap?" then follow up with "now execute it" and "show me how to handle the error I got" — all in the same session.
  • Ask about trade-offs. Questions like "What's the best way to handle large swaps?" or "When should I use limit orders instead of swaps?" get you a tailored recommendation backed by the documentation.
  • Ask for examples when building features. When you need working code, ask for an SDK example — the examples are tested, complete implementations that give the AI better context to help you.
  • Iterate on the code. Once the assistant retrieves an example, ask follow-up questions: "Now adapt this for USDT instead of USDC" or "Add retry logic for failed transactions."

Use of the 1inch MCP Server is subject to the 1inch Business Portal Terms of Service and the applicable Software Legal Notice. Users are responsible for ensuring compliance with applicable regulations.

Did you find what you need?