Technology in motion
Frixion

FRIXION

Managed infrastructure for autonomous AI agents.

One-command deployment for OpenClaw and Hermes. Persistent memory. Native wallets and x402 payments. Compute that bills only when your agent is awake.

$ frixion deploy --framework openclaw --tier standard
FX-001
v 1.0.0
What Frixion is

Infrastructure built
for the shape of agents.

AI agents idle for hours, then burst into sustained LLM calls, tool invocations and file operations. They need memory that survives restarts. They sometimes spin up browsers, execute untrusted code, or run long background jobs. Frixion's compute layer is built for that workload — not for stateless web apps.

Compute layer

Persistent compute

Sandboxed Linux runtimes per agent. Filesystems, memory, skills and conversation history survive restarts, tier changes and framework upgrades.

Orchestration

Auto deployment

One command provisions a sandbox, installs the framework, generates a wallet and registers the agent. Typical time from deploy to ready: under two minutes.

Payments

Native wallets

Every agent gets a wallet at deployment. Outbound x402 payments and inbound revenue, with per-transaction, per-day and domain-allowlist safety.

Channels

Channel gateway

Telegram, Discord, Slack, WhatsApp, Signal, iMessage, Matrix, Email, SMS — managed for you. Connect a channel, the gateway routes the message.

Security

Hard isolation

Tenant, project and agent boundaries are hard. Secrets are injected at runtime, never written to persistent storage. Wallet keys live behind the wallet service.

Cost model

Bursty pricing

Idle agents cost only a tier base reservation. Active compute bills per second. Most personal agents idle 90%+ of the time.

Frameworks

Two stable
deployment targets.

Pick OpenClaw for proactive, messaging-first agents that live in your apps. Pick Hermes for agents that learn over time and accumulate skills. Both deploy with one command. Both are model- agnostic. Both share the same channel gateway.

Stable

OpenClaw

Local-first gateway. Heartbeat-driven. Messaging-native.

An open-source personal agent built around a heartbeat scheduler that keeps the agent proactive — waking itself up on a configured cadence rather than waiting to be pinged. Personality and behaviour are defined in readable Markdown files.

Lives in WhatsApp, Telegram, Slack, Discord, iMessage
Heartbeat scheduler with configurable interval
Markdown-first configuration: SOUL.md, AGENTS.md, HEARTBEAT.md
Skills as Markdown files under skills/
Best for
Messaging-first personal assistants
Stable

Hermes

Self-improving. Skills library. Cross-session recall.

An open-source self-improving agent from Nous Research. Hermes autonomously creates skills from experience, refines them in use, and recalls past sessions through full-text search over its own history.

Learning loop — skills extracted from completed tasks
SQLite FTS5 index over all prior sessions
First-class MCP client; can also run as MCP server
Sub-agents with isolated conversation and terminal
Best for
Agents that get better the longer you use them
On the roadmap
Planned framework support: Letta, LangGraph runtime, Graphana
Coming soon
Wallets · x402 · Payments

Autonomous agents
need autonomous payments.

Every Frixion agent gets a wallet at deployment. Native support for the x402 protocol means your agent can pay for any x402- enabled resource — and serve as one. Per-transaction, per-day and domain allowlists make giving an agent a wallet reasonable.

x402 transaction flow
01
AGENTGET /api/research
02
SERVER402 + payment requirements
03
WALLETSign EIP-3009 payload
04
RETRYGET /api/research + payment
05
RESPONSE200 OK · settled in ~2s
Settled$0.0500 USDC · Base
x402 protocol

HTTP 402, revived.

Built by Coinbase for machine-to-machine commerce. Your agent hits a paid endpoint, the server returns 402, the wallet signs a payment payload, and the request retries with payment attached — all in roughly two seconds.

NetworksBase · Solana
TokenUSDC (EIP-3009)
Settlement~2 seconds
Inbound revenueFree of charge
Custody

Wallet keys held separately

The agent runtime can request signed transactions but cannot read or export private keys. A compromised agent can spend up to your limits — not beyond.

Limits

Per-tx, per-hour, per-day, per-month

Plus domain allowlists and an approval flow above a configurable threshold. Limits are the safety net that makes wallet-bound agents practical.

Roadmap

Credit cards in Q2 2026

Unified audit trail for fiat and crypto, per-agent virtual cards, and traditional SaaS subscriptions paid by your agent without crypto custody.

Pricing

Pay for the work,
not the idle.

Three things are billed separately: compute, payments and bandwidth. Compute splits into a small base reservation and per- second active time. Payments infrastructure is on us. Bandwidth gets 100 GB free per agent per month.

Starter

$5/ mo base
+ $0.001 / hour active compute
Personal agents and prototyping
1 vCPU · 8 GB RAM · 500 GB storage
Persistent workspace · 6h auto-snapshots
All channels available

Standard

Most popular
$20/ mo base
+ $0.001 / hour active compute
Daily-driver assistants in production
2 vCPU · 16 GB RAM · 1 TB storage
Browser automation · multi-channel
Priority deployment queue

Pro

$100/ mo base
+ $0.001 / hour active compute
Heavy skill libraries · multi-session
4 vCPU · 64 GB RAM · 5 TB storage
Frequent browser automation
Higher default rate limits

Need dedicated hosts, private networking or GPU access?

Contact sales
Dashboard

Operate every agent
from one console.

Project switcher, fleet view, per-agent observability, secrets, wallets, channels and billing — wired into a single wallet-gated workspace.

frixion.tech / app / overview
Project / Overview
Personal Assistant
Live
Agents
3
24h messages
532
Wallet (sum)
$235.14
Uptime 7d
99.93%
Agents3 running
atlas
OpenClaw
Standard
Running
orion-research
Hermes
Pro
Running
signal-ops
OpenClaw
Starter
Idle
FAQ

Common questions.

For anything not covered here, the full docs are bundled into the app or you can reach the team at support@frixion.xyz.

Managed infrastructure for AI agents: compute to host them, one-command deployment for popular frameworks, and wallets so they can pay for what they use.
Ready

The future, illuminated.

Connect your wallet and enter the dashboard. Deploy your first agent in under two minutes.