ClawRoam is a warm container handoff protocol for AI agent pods — inspired by IS-41, the telecom standard that gave cell phones seamless roaming in 1988. Your agent migrates between nodes with zero session gap.
We adapted the same prepare-before-you-cut pattern for AI agent pods. The target node is warm and ready before the source ever releases.
Roam controller monitors node load, health, and resource pressure. Detects when an agent should migrate.
Target node receives HANDREQ, pulls the agent's soul from Keeper (HLR), and starts warming the pod — while source is still active.
Target signals HANDREQ2 — "ready." Roam controller executes atomic routing cutover. Source drains. Zero perceived gap.
Agent is running on the target with full soul state. Sessions continue. The user never noticed it moved.
Modeled directly on IS-41. Every message has a telecom analog that's been proven at carrier scale for 35 years.
Agent pod registers with Keeper (HLR) and local clawtermd (VLR) on startup. Establishes soul_id, soul_version, and socket path.
Controller sends HANDREQ to target node. Target warms the pod, pulls soul from HLR, signals HANDREQ2 when ready. Source stays active throughout.
Controller sends HANDEXEC to source node. Routing table atomically updated. Sessions redirect to target. Source enters drain mode.
Source node deregisters from Keeper after drain completes. Pod terminates. VLR cache purged. HLR reflects new home node.
Every IS-41 entity maps directly to a ClawRoam component. If you know SS7, you already understand this.
ClawRoam is in active development as part of the Claw Network Architecture. Early access for builders on the waitlist.
Part of the Claw Network Architecture — ClawTerm · ClawCell · ClawRoam · ClawGuardian