Pricing

OpenClaw pricing for useful, safe agent work.

Setup gets the environment running. Build work turns it into a working agent stack. Retainer support keeps the system useful, reviewed and improving after launch.

Packages

Three sensible starting points.

The numbers below are designed around scoped work, not vague AI advisory time. Each package assumes a clear workflow, documented controls and practical handover.

01 · Setup

OpenClaw Setup

£3,500

one-off · 2–3 weeks

A properly structured OpenClaw deployment your team can actually operate. Workspace, memory, core skills, channels, docs, and training in one focused setup project.

Best for getting from zero to a safe, working agent.

  • Workspace + role architecture
  • Memory + knowledge structure
  • 5 custom starter skills
  • Tool + channel integrations
  • Operating runbook
  • Team training session
Start setup
Most chosen

02 · Build

Agentic Stack

£6,500

project · 4–6 weeks

Move from one assistant to a coordinated agent team. Orchestration, security boundaries, model routing, evals, and production operating rules.

Best for teams moving from one assistant to coordinated workflows.

  • Multi-agent workflow design
  • Model + cost routing
  • Sandbox + approval boundaries
  • Eval pipeline + monitoring
  • Production operating rules
  • Handover to retainer
Scope a build

03 · Operate

Retainer

£950 / mo

rolling · 3-mo minimum

The setup is the start. Retainer covers eval reviews, prompt updates, skill changes, model migrations, troubleshooting, and continued capability development.

Best for keeping live agents useful, safe and improving month by month.

  • Monthly evals + audit
  • Prompt + skill updates
  • Model migration support
  • New skills (2 / month)
  • Priority response
  • Quarterly review
Add retainer

Free OpenClaw resource

Safe setup checklist and permission matrix.

Use this before giving an agent access to files, inboxes, CRMs, production systems, or client data. It turns vague AI risk into a plain set of permissions, approvals and evidence.

Map the workflow before choosing tools
Decide read, draft, approve and execute permissions
Set approval gates for external messages and risky changes
Log agent actions, sources, owner decisions and rollback steps

How to budget

Do not buy an agent. Buy a controlled workflow.

The safest first budget is the smallest budget that proves one real process can be improved without handing the agent more access than the job needs.

Start with a narrow workflow

The right first project has one owner, a visible input, a useful output and a reviewable result.

Price the controls

Permissions, approvals, logs, documentation and handover are part of the build, not optional polish.

Expand after evidence

The second workflow should be funded by what the first workflow proved, not by a broad AI roadmap.

Scope a setup

Send the workflow. I will tell you the sensible first budget.

Include the systems involved, who owns the decision, and where the agent must ask before acting. That is enough to turn a rough idea into a useful setup scope.

Fallback form only. The fastest route is the discovery call.