For developers
Durable runs with status, events, artifacts, retries, and audit history — submit once, inspect anytime.
Daemon Loom Cloud
Durable AI workflows
in the background
Most tools help agents think or call tools. Daemon Loom Cloud owns what makes long-running work dependable: typed handoffs, watchdogs, human gates, and traceable state. Submit once — get runs, events, artifacts, approvals, and audit history.
Built for every role in the loop
For agent clients
MCP tools and resources for IDEs and assistants — submit workflows, list runs, read state, approve gates.
For operators
Live workflows, human gates, policy blocks, stalled packets, and finalized artifacts in one control plane.
Primary surfaces
One operations plane — many ways to submit, inspect, and integrate.
| Surface | Audience | Purpose |
|---|---|---|
| REST / gRPC API | Platform teams | Submit, validate, run, pause, resume, cancel, inspect, approve |
| MCP server | IDEs, assistants | Submit workflows, list runs, read state, approve gates |
| Admin console | Customer admins | Run timeline, packet graph, approval queue, usage |
| Webhooks / streams | Customer systems | Run events, gate requests, artifact ready, finalization |
| SDKs / CLI | Developers | Typed clients, validation, run tailing, artifact download |
Workflow drop-off
Author a workflow in YAML or via API — easier than assembling an ad hoc agent loop. The loom turns it into auditable packets, runs specialists under persistent daemons, and returns events and artifacts on your schedule.
- Goal and inputs — what the workflow should accomplish
- Outputs and artifacts — what you get back when it finishes
- Schedule — one-shot or recurring cadence
- Approvals — human gates before sensitive actions
- Policies — guardrails enforced by the control plane
workflow:
name: weekly-competitive-intel
goal: Research competitors and publish an executive summary every Monday.
schedule:
mode: recurring
cadence: weekly
approvals:
require_human_before:
- external_publishD-Loom under the hood
Daemon Loom Cloud runs on the same D-Loom discipline as daeloom.ai: disposable Needles, a Pattern Ledger as source of truth, and process-completion TTL instead of arbitrary wall-clock timeouts. Same runtime discipline — surfaced as a hosted product with API, MCP, and operator console.
Under the hood → D-Loom execution modelClusterClaw hosting
Workloads run on ClusterClaw isolation lanes — Cloud Run by default, optional GKE for enterprise. BYOK, managed OpenClaw gateways, and honest infra boundaries: your workflows get durable orchestration without building the substrate yourself.
Cloud Run default
Per-tenant isolation lanes with managed scaling — the default path for cloud beta.
GKE enterprise
Optional dedicated clusters for customers who need stronger isolation and custom networking.
Pricing & availability
Daemon Loom Cloud is in beta — contact us for developer access. No fabricated pricing tiers yet; API docs, MCP endpoints, and a status page are on the roadmap.
Join the developer waitlistClusterClaw home · daeloom.ai · API docs · MCP · status — coming soon