BETA
Sui

Chain

Valseq MCP

The missing context.

Your model can figure out a lot on its own. Valseq adds the context that’s missing: why a transaction failed, the route it took and who else was there.

Claude

https://valseq.xyz/mcp

In Claude, open Customize → Connectors → Add custom connector. Name it Valseq, paste the endpoint and choose No sign in. Enable Valseq through + → Connectors in your conversation. Your workspace owner may need to add it first.

Claude Code CLI

claude mcp add --scope user --transport http valseq https://valseq.xyz/mcp

Run it once in your terminal. The user scope makes Valseq available in all your projects.

ChatGPT

https://valseq.xyz/mcp

In ChatGPT on the web, open Plugins → + → Add custom MCP server. Name it Valseq, paste the endpoint and choose No authentication. Create and install the plugin, then select @Valseq in a Work conversation. Availability depends on your account and workspace settings.

Codex CLI

codex mcp add valseq --url https://valseq.xyz/mcp

Run this command in a terminal with Codex installed. Start a new session and use /mcp to check the connection.

Others

https://valseq.xyz/mcp

Add a remote server named Valseq in your MCP client. Choose Streamable HTTP, paste the endpoint and leave authentication empty. Enable its tools before sending your first question.

Read-only, no API key.

An abort code is not an answer.

One failed trade, as a Sui node returns it and as your agent explains it with Valseq.

Raw node response, excerpt
{
  "effects": {
    "status": { "status": "failure", "error": "MoveAbort(MoveLocation { module: ModuleId { address: 90290f…2df56, name: Identifier(\"flash_loan\") }, function: 4, instruction: 11, function_name: Some(\"return_flashloan_quote\") }, 14680251) in command 6" },
    "gasUsed": { "computationCost": "1338250", "storageCost": "19402800", "storageRebate": "19400860", "nonRefundableStorageFee": "1940" }
  },
  "events": [],
  "balanceChanges": [{ "amount": "-1340190", "coinType": "0x2::sui::SUI" }]
}

An abort code, no events and a gas charge. Nothing about why it failed, what ran or who else was there.

Your agent with Valseq

Why did 7Mjivs…t39n fail? What else used the same pools?

  1. valseq get_transaction(view: "brief")
  2. valseq get_transaction_context(focus: "objects")

The transaction borrowed 204.604792767 SUI from DeepBook and swapped it through Cetus and Turbos. It ended with 204.370735062 SUI, leaving it 0.234057705 SUI short of repaying the loan.

The repayment check failed, so all swaps were reversed. The sender paid 0.00134019 SUI in gas.

Across three checkpoints, 27 transactions from 10 senders used the same Turbos pool. Only one succeeded.

We tested the same agent with and without Valseq MCP.

Benchmarked gpt-6.1-sol with and without Valseq MCP 1.9.4. AI-graded coverage on selected Sui investigations.

  • Required facts covered

    Without Valseq
    79%
    With Valseq
    95%

    Share of the answer key’s required facts the answers contained.

  • Caveats covered

    Without Valseq
    81%
    With Valseq
    98%

    Share of the required caveats the answers stated.

  • Tool calls

    Without Valseq
    8.5
    With Valseq
    4

    Median shell commands and Valseq calls per investigation.

  • Tokens

    Without Valseq
    275.2K
    With Valseq
    119.2K

    Median client tokens, including cached context; not billed usage.

Context your model can check.

  1. Transactions

    get_transaction

    Replayed and checked against the chain, with DeFi operations decoded from their published signatures.

  2. Related activity

    get_transaction_context

    Nearby transactions that share arguments, objects or event values, within a stated window.

  3. History and code

    get_object·fetch

    Exact object versions before and after, and the published Move code behind them.

  4. Network research

    query_transactions

    Transaction counts, gas and balance changes across checkpoints, with stated coverage.

Every finding is labeled

Proven
Read directly from recorded chain data.
Derived
Worked out from proven evidence, like the direction of a trade.
Unknown
Missing from the sources. Reported as unknown, never as zero.

Evidence, not guesses.

Add Valseq to your agent.

Claude

https://valseq.xyz/mcp

In Claude, open Customize → Connectors → Add custom connector. Name it Valseq, paste the endpoint and choose No sign in. Enable Valseq through + → Connectors in your conversation. Your workspace owner may need to add it first.

Claude Code CLI

claude mcp add --scope user --transport http valseq https://valseq.xyz/mcp

Run it once in your terminal. The user scope makes Valseq available in all your projects.

ChatGPT

https://valseq.xyz/mcp

In ChatGPT on the web, open Plugins → + → Add custom MCP server. Name it Valseq, paste the endpoint and choose No authentication. Create and install the plugin, then select @Valseq in a Work conversation. Availability depends on your account and workspace settings.

Codex CLI

codex mcp add valseq --url https://valseq.xyz/mcp

Run this command in a terminal with Codex installed. Start a new session and use /mcp to check the connection.

Others

https://valseq.xyz/mcp

Add a remote server named Valseq in your MCP client. Choose Streamable HTTP, paste the endpoint and leave authentication empty. Enable its tools before sending your first question.