Your agent
The Peck Harness runs on the node and is reachable only by its owner. It is Peck's own agent harness, running on hardware you control.
early access node #1 is ours
A node is a small box you own. It runs your agent, reads the BSV chain straight from peer nodes, and serves a small API, all behind one URL that you open with your BRC-100 wallet. No passwords, no accounts.
Asking node #1 for its status…
Read live from mini.peck.run/api/v1/status by your browser. Nothing is cached or proxied here.
Signed offers listed by the peck.network registry at pulse.peck.host. Uptime is measured by independent checks, not claimed by the node.
| Node | Access | Price | Uptime 24 h |
|---|
The registry is reachable but lists no live offers right now.
An agent that acts for you should run where you can see it, read the chain from a source you chose, and answer to a key only you hold. A node is that, in one box.
The Peck Harness runs on the node and is reachable only by its owner. It is Peck's own agent harness, running on hardware you control.
The node listens to BSV peers directly: block headers and Peck transactions. There is no block explorer API in between.
Public, read-only and open to any origin: health, status, chain headers and a live feed of Peck transactions as they hit the mempool.
You log in by signing a challenge with your BRC-100 wallet. The first login with the one-time claim code makes you the owner.
| Part | Job | Memory budget |
|---|---|---|
| gate (Go) | Node page, public API, wallet login, forward-auth for the Harness, live mempool feed, signed offers to peck.network. | about 20 MB |
| peck-gossip (Go) | BSV peer connections for headers and Peck transactions. | about 50 MB |
| Peck Harness (Node) | Your agent. Calls a model through llm.peck.to with a metered key. | 200 MB to 1.5 GB |
| Caddy | TLS in front of the gate and the Harness. | about 40 MB |
Measured, not budgeted: on 7 Oct 2026, gossip and the Harness together sat at about 150 MB while idle on node #1. The table above is the budget we plan against, not a measurement.
Node #1 hardware: a 2018 Mac mini (Intel i3, 4 cores, 7 GB RAM) in our office that also does other jobs. So every part of the node runs in a systemd slice capped at 3 GB of memory with low CPU priority, as its own user without sudo, with no access to home directories and no route to private networks.
Early access. The node software is private while it is hardened, so there is no public download yet. If you want a node, tell us and we will work out access with you.
On a build machine, deploy/push-artifacts.sh you@your-node builds the gate and gossip, bundles Node, the Peck Harness and Caddy, and copies them to the node over SSH.
On the node, sudo deploy/install.sh creates the sandboxed users, the systemd units, the firewall and the Caddy config, and generates the node key. Running it again keeps your keys and claim state.
Print the one-time claim code on the box. Open https://your-node/login, connect your wallet and enter the code. Your wallet signs twice: the login, then an owner attestation that peck.network and peck.run can check.
What you need: a Linux machine with systemd (node #1 runs Ubuntu), a public name with TLS, outbound access to BSV peers, a BRC-100 wallet such as BSV Desktop or Peck OS, and a model key for the Harness. The installer has only been exercised on node #1 so far. Nodes without a public IP need a relay from peck.run, which is phase 3 and not built.
Phase 2. Not live. Today every service on node #1 is free or owner-only. Nobody pays a node, and no node earns anything yet.
The plan is that someone who is not the owner pays your node per hour or per call, through a payment channel (peck.channel) or a plain BRC-29 payment, and the Harness learns to serve more than one person. Payment channels stay switched off on peck.run until channel funding is verified on-chain, so paid use waits for that.