Early access

An MCP server that simulates and risk-checks every transaction an AI agent wants to sign, before it signs.

It works at the transaction level for any agent framework, rather than being another agent wallet. It returns machine-readable reasons so the agent can adjust and retry, not just fail.

Core features
5
Status
Early access
Network
Solana
Your funds
Non-custodial
MCP server exposing a check_transaction tool that simulates a transaction and returns balance changes, programs touchedToken checks: mint and freeze authority, top holder concentration, liquidity depth and slippage versus the requestedRules file in plain YAML (max size per trade, allowed programs, blocked tokens) with the reason for each verdictPublic verdict log with shareable permalinks so a builder can show why a trade was blockedWeb playground where anyone can paste a transaction and see the decoded result
Preflight control
OwnerSets rules
AgentProposes
PolicyChecks
ExecutionSettles
Recent runs
AgentActionAmountResult
Dope agentSwap within limit$4.4KNeeds approval
Rifa agentSwap within limit$4.7KExecuted
Nepe agentClaim rewards$450Executed

Spending limit

$1.3K of $10.0K used today

Policies

Allowlisted programs only
Approval above $2,500

Approvals

  • Swap within limitwaiting
Features

What you get with Preflight.

Built for builders of Solana trading and portfolio agents, funds running automated strategies, and power users who let agents prepare transactions for them.

  • MCP server exposing a check_transaction tool that simulates a transaction and returns balance changes, programs touched

    Planned
  • Token checks: mint and freeze authority, top holder concentration, liquidity depth and slippage versus the requested

    Planned
  • Rules file in plain YAML (max size per trade, allowed programs, blocked tokens) with the reason for each verdict

    Planned
  • Public verdict log with shareable permalinks so a builder can show why a trade was blocked

    Planned
  • Web playground where anyone can paste a transaction and see the decoded result

    Planned
Activity

Token checks: mint and freeze authority, top holder concentration, liquidity depth and slippage versus the requested.

Every proposal, check and result in one log. Filter by agent, see why something was blocked, and export it for review.

Planned
Activity
AgentActionAmountResult
Dope agentSwap within limit$4.4KNeeds approval
Rifa agentSwap within limit$4.7KExecuted
Nepe agentClaim rewards$450Executed
Ripe agentPay invoice$4.1KExecuted
Tomor agentBridge request$4.6KExecuted
Sado agentBridge request$2.2KExecuted
Policies

MCP server exposing a check_transaction tool that simulates a transaction and returns balance changes, programs touched.

Write the rules once, in plain settings. Every action an agent proposes is checked against them before anything is signed.

Planned

Policy: treasury agent

active
Allowlisted programs only
Require approval above $2,500
Block new counterparties
Pause outside 08:00 to 22:00
Daily limit$10,000
Controls

Rules file in plain YAML (max size per trade, allowed programs, blocked tokens) with the reason for each verdict.

Stop one agent or all of them. Pausing takes effect before the next proposal, and nothing queued is executed while paused.

Planned

Execution

ProposeAgent submits an intent
CheckPolicies run
SettleSigned and sent

It works at the transaction level for any agent framework, rather than being another agent wallet. It returns machine-readable reasons so the agent can adjust and retry, not just fail.

How it works

4 steps from start to first result.

Connect an accountOwners keep the keys.
Set policiesLimits, programs, approval thresholds.
Give agents scoped accessRevocable at any time.
Review the logEvery action is recorded.
Under the hood

Open by default.

Every screen reads from the same model. Live parts read public onchain data; nothing ever asks your wallet to sign.

Read the docs
preflight.json
{
  "product": "Preflight",
  "network": "solana",
  "features": [
    {
      "name": "MCP server exposing a check_transaction tool tha",
      "status": "planned"
    },
    {
      "name": "Token checks: mint and freeze authority, top hol",
      "status": "planned"
    },
    {
      "name": "Rules file in plain YAML (max size per trade, al",
      "status": "planned"
    },
    {
      "name": "Public verdict log with shareable permalinks so ",
      "status": "planned"
    }
  ]
}

What works today.

Live features work today. Planned ones are next on the roadmap, in this order.

MCP server exposing a check_transaction tool that simulates a transaction and returns balance changes, programs touched.Planned
Token checks: mint and freeze authority, top holder concentration, liquidity depth and slippage versus the requested.Planned
Rules file in plain YAML (max size per trade, allowed programs, blocked tokens) with the reason for each verdict.Planned
Public verdict log with shareable permalinks so a builder can show why a trade was blockedPlanned
Web playground where anyone can paste a transaction and see the decoded resultPlanned

Questions.

Short answers. The docs have the rest.

What is Preflight?

An MCP server that simulates and risk-checks every transaction an AI agent wants to sign, before it signs. It works at the transaction level for any agent framework, rather than being another agent wallet. It returns machine-readable reasons so the agent can adjust and retry, not just fail.

What works today?

Preflight is in development. Every feature is on the status list with what ships next, and the app opens in a sample workspace so you can see how it will work.

Do I need a wallet?

No. You can explore without one. Connecting a Solana wallet only shares your public address so Preflight can read your balances; it never asks you to sign a transaction.

Who is it for?

Builders of Solana trading and portfolio agents, funds running automated strategies, and power users who let agents prepare transactions for them.

MCP server exposing a check_transaction tool that simulates a transaction and returns balance changes, programs touched.

An MCP server that simulates and risk-checks every transaction an AI agent wants to sign, before it signs.