AstraFlow

Agentic Solana Tools

Live on SolanaPolicy-first execution

The agent that watches Solana — and can execute safely.

Wallet tracking, token launch monitoring, and automated trades with strict guardrails. Chat to configure. Tools to execute.

Avg. preflight

< 250ms

Tool calls

Guarded

Keys

Never in LLM

* Trading involves risk. This project is designed for safety and transparency, but you control the limits.

terminal

quickstart

Install + onboard

$ curl -fsSL https://www.astraflow.fun/install.sh | bash
$ astraflow onboard
$ astraflow watch --wallet 7nQ...k3p

Wallet tracking

Live swaps, transfers, and alerts.

Separate signer

Execution service holds keys.

Risk guard

Limits + allowlists by default.

Fast routes

Quotes + swaps via aggregator.

Try it

Interactive command playground

Click-run a safe demo flow. No real trading — just shows how the agent plans, checks policy, simulates, and executes.

Controls

Pick a mode, set a wallet, and run the flow.

Wallet

Mode

Safe mode

Extra checks + stricter confirmations.

Tip: wire this to your backend by replacing the demo runner with your API call.

Output

A rolling log window (like a bot console).

> agent.status

OK · connected to Solana RPC

Policy: risk-guard enabled

Architecture

Chat to configure. Tools to execute.

A clean separation between intent, policy, data, and signing — so the system stays controllable.

Clickable agent flow

Tap a node to see what it does.

Chat Intent

active

You describe goals in plain language (track wallets, alerts, safe auto-trades).

Deterministic execution through tools

Preflight simulation before signing

Full audit logs for every decision

Track wallets in real time

Subscribe to swaps/transfers, build your own watchlists, and trigger alerts or automation rules.

Tx parsing + classification

Custom filters + cooldowns

Alert routing (chat/webhooks)

Automate early entries — with guardrails

Monitor liquidity and route availability, then place small capped entries only when checks pass.

Min-liquidity thresholds

Allowlist programs/DEX

Strict slippage and sizing

Simulate first, then sign

Separate the AI from the signer. The executor simulates, signs, and broadcasts — safely.

LLM never sees private keys

Executor service isolation

Retries + confirmation tracking

Composable rules

Copy-trade, alerts, or custom strategies — all as configurable policies.

Policy engine

Hard stops for slippage, size, frequency, and program allowlists.

Observability

PnL, positions, confirmations, and an audit trail for every tool call.

Works with your stack

Mix RPC providers, aggregators, and chat surfaces.

JupiterRaydiumOrcaHeliusQuickNodeTelegramDiscordWebhooks

Security

Designed so the AI can’t go rogue

The agent can propose actions, but the policy layer and executor decide what actually happens.

Three-layer safety model

What protects you when markets get weird.

Policy & risk limits

Max position size, slippage caps, frequency limits, allowlists, confirmations.

Simulation & sanity checks

Preflight simulation, route validation, balance checks, rejection on mismatch.

Isolated signing

Keys live in the executor (vault/HSM/encrypted secret). The LLM never touches secrets.

Execution modes

Choose the level of autonomy you want.

ManualConfirmAuto

Start with Manual, move to Confirm when your rules are stable, then enable Auto only for small, capped strategies.

Safety checklist

Default guardrails you should keep on.

Simulate before every send

Max slippage + max spend

Program allowlist (DEX + token)

Daily loss cap + cooldowns

Audit logs for every action

You can tighten these further per strategy (e.g., min-liquidity and confirmation rules).

Pricing

Start small. Scale when your rules work.

Simple plans for builders and teams. Swap providers and infrastructure as you grow.

Builder

$29

per month

1 executor

Wallet tracking + alerts

Manual/Confirm modes

Community templates

Cancel anytime. Switch plans whenever.

Questions, answered

If you’re building an executing agent, these are the things that matter.

FAQ

Short + practical.

No. Keys live in an isolated executor service (vault/HSM/encrypted secret). The AI can only request tool actions that the policy engine approves.

Stay in the loop

Get updates on features, integrations, and security improvements.

No spam. Unsubscribe anytime.

Security-first by default

This page is a starting point. Your production build should include monitoring, rate limits, and strict policy enforcement.