The OpenClaw Dashboard is the SFAI Labs control plane for teams running more than one OpenClaw agent. It consolidates live session activity, spend, cron runs, memory state, approvals, and audit evidence into a single pane of glass, and it is designed to do what local JSONL logs cannot: give a platform team real-time visibility and real policy control across a fleet.
This page explains what the OpenClaw Dashboard is, the problem it solves, the capabilities shipping today, the persona-based views for engineering, finance, and compliance, and how to get a demo.
The Problem: Fleets Without Visibility
OpenClaw is easy to run with one agent. You read the gateway logs. You check the session transcripts on disk. You tail the cron output. When something breaks, you open a shell and grep.
That workflow stops scaling around agent number three. A platform team running ten OpenClaw agents across five customers or business units faces a different problem:
- No single place to see what every agent is doing right now. Live sessions live on different machines. Cron state lives in per-agent directories. Sub-agent runs are scattered across log files.
- No aggregate spend view. Each agent reports its own token usage. Rolling up spend across the fleet means writing a script.
- No policy layer. An agent can burn a budget, touch a sensitive system, or run an unapproved action because nothing is standing between the model and the tool.
- No approval workflow. When an agent needs a human to sign off on a sensitive operation, there is no inbox. Approvals happen over Slack or not at all.
- No audit evidence. Local JSONL logs are deletable. Compliance teams cannot use them as-is for SOC 2 or ISO 27001 walkthroughs. See our audit logging guide for the full gap.
The OpenClaw Dashboard is built to close these five gaps at the same time.
What the OpenClaw Dashboard Gives You
The dashboard exposes five capabilities. Each one maps to a problem above.
Agent Observability
A live fleet view shows every OpenClaw agent connected to the dashboard: which model it is running, whether the heartbeat has fired on schedule, what the last session did, and whether any sub-agents are active. Click into an agent and the session tree expands. You see the user prompt, the reasoning trace, every tool call, and the final response.
Traces are designed to work with OpenClaw’s existing OpenTelemetry output, so the dashboard does not require a parallel instrumentation layer. If OpenClaw already emits traces in your environment, the dashboard reads them.
The observability view answers the question a platform lead asks fifty times a day: what is this agent doing right now, and did it do what I asked?
Policy Controls
Observability shows what happened. Policy stops what should not happen.
The dashboard is built to let platform teams define guardrails that run before a tool call executes:
- Resource allowlists. An agent can only touch the integrations you explicitly approve.
- Action thresholds. A write operation above a configured size or value requires approval.
- Kill-switches. One click pauses an agent or the whole fleet if something is going wrong.
Policy is the capability most OpenClaw observability tools skip. Shipping policy as a first-class concern is the reason a platform team deploys the SFAI Labs Dashboard instead of rolling their own Grafana.
Spend Tracking
Token costs are the first thing any engineering lead running a fleet wants to see. The dashboard aggregates token and dollar cost per agent, per model, per session, and per customer or tag. A daily and monthly roll-up sits on the home view.
Budgets sit on top of the spend view. A per-agent daily cap, a per-customer monthly cap, or a fleet-wide ceiling can be configured. When a budget is hit, the agent pauses and the owner is notified. No silent overruns. No surprise invoices at the end of the month.
Human-in-the-Loop Approvals
Some agent actions should never happen without a human in the loop. Sending an external email to a customer. Paying an invoice. Running a destructive database operation. Posting a public statement.
The dashboard provides an approval inbox for exactly these cases. When an agent hits a guarded action, it pauses and sends an approval request to the queue. A reviewer sees the intended action, the reasoning that led to it, and the upstream context. One click approves or rejects. The agent resumes or rolls back.
This is the feature that lets a platform team deploy OpenClaw against workflows that would otherwise require a full human operator.
Audit Log
The dashboard ships a central, append-only audit log that aggregates session transcripts, gateway events, cron runs, policy decisions, and approval outcomes across every connected agent.
The audit log is built for compliance evidence: role-mapped retention, export to your SIEM, and query filters that map to SOC 2 Common Criteria and ISO 27001 Annex A controls. The full control mapping is in the OpenClaw audit logging guide.
If you are taking OpenClaw into an enterprise deal that requires a SOC 2 walkthrough, this is the evidence layer the auditor will ask for.
See it live — request a demo →
How It Connects to OpenClaw
The dashboard is a separate service that connects to every OpenClaw agent in your fleet. The connection uses the same gateway mechanism OpenClaw already exposes for multi-user setups and remote control.
At a high level:
- OpenClaw agents run where they already run: on-prem, on a VPS, on a Mac mini, or inside a customer VPC.
- The dashboard service runs either as an SFAI Labs managed instance or as a self-hosted deployment inside your own infrastructure.
- The connection is outbound from each agent to the dashboard, authenticated with a gateway token.
This design matters for two reasons. First, agents keep running if the dashboard is unreachable — the dashboard is a control plane, not a dependency. Second, self-hosted deployments keep agent data inside your boundary for teams that cannot send workloads to a SaaS.
For the gateway mechanism itself, see the OpenClaw gateway token guide.
Who It Is For
The OpenClaw Dashboard is built for four overlapping personas inside a company running OpenClaw at scale.
Platform Engineering Leads
You are the person on-call when an agent breaks. You need a trace view, a kill-switch, and a reliable way to answer “is the fleet healthy?” without grep. The dashboard is built to cut the time from alert to root cause to minutes.
CTOs and VPs of Engineering
You own the spend and the risk. You need to see aggregate cost trends, approval volume, and incident counts in a weekly-review view. The dashboard is designed so you never have to ask the platform team to compile a fleet report by hand.
Compliance and Security
You need evidence. The audit log, role-based access, approval records, and policy decisions are the artifacts your auditor will pull. The dashboard is built to produce this evidence as a byproduct of normal operation, not as a scramble before an audit.
Finance and Operations
You sign the bill. The spend view, budget caps, and per-customer cost rollups let you forecast and chargeback accurately. If your agents serve internal teams or paying customers, this is the view that keeps the economics honest.
Pricing and How to Get Started
Pricing is based on fleet size and deployment model (managed vs. self-hosted). SFAI Labs does not publish a public price list because early deployments are scoped per customer.
The fastest path:
- Email the team with the number of agents you run and whether you need self-hosted or managed.
- Demo call — 30 minutes. We show the live dashboard against a fleet of test agents and discuss your setup.
- Pilot — most pilots start with a single-agent instrumentation, expand to fleet, then add policy and approval rules.
Frequently Asked Questions
What is the OpenClaw Dashboard?
The OpenClaw Dashboard is a control plane for teams running multiple OpenClaw agents. It provides live observability, policy controls, spend tracking, human-in-the-loop approvals, and a compliance-grade audit log across a fleet.
Is the OpenClaw Dashboard open source?
No. The dashboard is a commercial product from SFAI Labs. OpenClaw itself remains open source and the dashboard connects to unmodified OpenClaw agents through the standard gateway.
How does it differ from ClawMetry or community dashboards?
Community dashboards focus on observability — showing what the agent did after the fact. The SFAI Labs Dashboard is built around policy and approvals: stopping sensitive actions before they run, not just reporting on them afterward. It also targets multi-agent fleets and compliance evidence rather than a single-user local view.
Can I self-host it?
Yes. Self-hosted deployments are supported for teams that cannot send agent data to a managed service. Self-hosting covers air-gapped and VPC-contained environments.
Does it require changes to my OpenClaw agents?
No code changes. Connection is through the existing OpenClaw gateway with a token. If your agents already run multi-user or remote setups, the instrumentation path is the same.
What data does the dashboard store?
Session transcripts, tool calls, cost records, cron run outcomes, policy decisions, and approval events. Data retention is configurable per workspace to match your compliance requirements.
Does it support SSO?
SSO integration is on the roadmap and tracked alongside our OpenClaw SSO integration work. Contact us for the current status.
Key Takeaways
- The OpenClaw Dashboard is the SFAI Labs control plane for multi-agent OpenClaw deployments.
- It is built to close five gaps that emerge past the first agent: live observability, policy before execution, spend control, human-in-the-loop approvals, and a compliance-grade audit log.
- It connects to unmodified OpenClaw agents through the existing gateway and supports managed or self-hosted deployment.
- Pricing is scoped per customer. The fastest path to an evaluation is a 30-minute demo.
SFAI Labs
