Dexra is a Solana data and risk-analysis API for AI agents and trading bots. One call returns a finished answer about a token launch, a token and its holders, a wallet, a liquidity pool, a perp market or a planned swap: a verdict, a plain-language summary, flags and the evidence behind them. Agents use it over MCP, bots over REST.

How do I connect Dexra to Claude, Cursor or another MCP client?

Add the MCP server at https://api.dexra.dev/mcp (Streamable HTTP) with the header "Authorization: Bearer ". In Claude Code: claude mcp add --transport http dexra https://api.dexra.dev/mcp --header "Authorization: Bearer dxr_…". Claude Desktop and Cursor take the same URL and header in their MCP config.

What does Dexra check that a token safety score does not?

Before measuring holder concentration Dexra merges wallets that were funded from the same address, so a team spread over many wallets counts as one holder. It maps who funded a launch’s first buyers, checks the creator’s previous launches and how many collapsed, measures sandwich-bot and bot share on a pool, compares all-in execution cost across perp venues, and checks blacklist proximity within two transfers.

Which launchpads, pools and venues are covered?

Launches on pump.fun, LetsBonk, Raydium LaunchLab, Meteora DBC, PumpSwap and Raydium AMM and CPMM; concentrated-liquidity pools on Raydium CLMM, Orca and Meteora DLMM; perp markets on Binance, Bybit, OKX, Hyperliquid, Drift and Jupiter.

How much does a call cost?

Calls are priced in credits; one credit is 0.005 USDC, the cheapest call. A launch check or swap preflight uses 1 credit and a deep token, holder, wallet or pool analysis 6. Free accounts get 100 credits a day once the email is confirmed; paid plans include more credits a day. Three tools are always free: describe_address, service_stats and score_performance.

Can an agent pay without an account?

Yes. Without an API key, paid tools answer with an x402 payment request; the agent pays the call in USDC on Solana and repeats it with the payment. A call that fails is not settled, so nothing leaves the wallet.

Does it work for trading bots without AI?

Yes. Every tool is also a REST endpoint (GET with query parameters, POST for batches) with the same arguments and answers, described in an OpenAPI 3.1 file at /openapi.json. Watches post a signed webhook when an answer changes, so a bot does not have to poll.

How fresh is the data?

Answers are computed from Solana on-chain data and venue APIs when you call. Every answer carries a generated_at timestamp and the sources and coverage of the history it used, so an agent can tell how much evidence stands behind a verdict.

Is a verdict financial advice?

No. Verdicts, scores and flags are automated readings of on-chain and market data. They can be incomplete or wrong; you and your agent decide what to trade.

FAQ · Dexra docs

FAQ

What is Dexra?: Dexra is a Solana data and risk-analysis API for AI agents and trading bots. One call returns a finished answer about a token launch, a token and its holders, a wallet, a liquidity pool, a perp market or a planned swap: a verdict, a plain-language summary, flags and the evidence behind them. Agents use it over MCP, bots over REST.
How do I connect Dexra to Claude, Cursor or another MCP client?: Add the MCP server at https://api.dexra.dev/mcp (Streamable HTTP) with the header "Authorization: Bearer <your API key>". In Claude Code: claude mcp add --transport http dexra https://api.dexra.dev/mcp --header "Authorization: Bearer dxr_…". Claude Desktop and Cursor take the same URL and header in their MCP config.
What does Dexra check that a token safety score does not?: Before measuring holder concentration Dexra merges wallets that were funded from the same address, so a team spread over many wallets counts as one holder. It maps who funded a launch’s first buyers, checks the creator’s previous launches and how many collapsed, measures sandwich-bot and bot share on a pool, compares all-in execution cost across perp venues, and checks blacklist proximity within two transfers.
Which launchpads, pools and venues are covered?: Launches on pump.fun, LetsBonk, Raydium LaunchLab, Meteora DBC, PumpSwap and Raydium AMM and CPMM; concentrated-liquidity pools on Raydium CLMM, Orca and Meteora DLMM; perp markets on Binance, Bybit, OKX, Hyperliquid, Drift and Jupiter.
How much does a call cost?: Calls are priced in credits; one credit is 0.005 USDC, the cheapest call. A launch check or swap preflight uses 1 credit and a deep token, holder, wallet or pool analysis 6. Free accounts get 100 credits a day once the email is confirmed; paid plans include more credits a day. Three tools are always free: describe_address, service_stats and score_performance.
Can an agent pay without an account?: Yes. Without an API key, paid tools answer with an x402 payment request; the agent pays the call in USDC on Solana and repeats it with the payment. A call that fails is not settled, so nothing leaves the wallet.
Does it work for trading bots without AI?: Yes. Every tool is also a REST endpoint (GET with query parameters, POST for batches) with the same arguments and answers, described in an OpenAPI 3.1 file at /openapi.json. Watches post a signed webhook when an answer changes, so a bot does not have to poll.
How fresh is the data?: Answers are computed from Solana on-chain data and venue APIs when you call. Every answer carries a generated_at timestamp and the sources and coverage of the history it used, so an agent can tell how much evidence stands behind a verdict.
Is a verdict financial advice?: No. Verdicts, scores and flags are automated readings of on-chain and market data. They can be incomplete or wrong; you and your agent decide what to trade.