Product 001 — Early access
SmartStaff
The agent meta harness.
One operating layer above every AI agent you run. SmartStaff gives agents shared memory, governed tools, policies and oversight, so a crowd of bots becomes a staff you can actually manage.
01 — The idea
A harness runs one agent. A meta harness runs the staff.
Harness
The scaffolding around one agent
Model, prompt, tools, context window. Every agent framework ships one, and every one is slightly different.
Meta harness
The layer above many harnesses
Shared memory, routing, policy, evaluation and audit, applied consistently to every agent regardless of what it was built with.
SmartStaff
Our meta harness, built for operators
Plug in the agents you already use or create new ones, then run them as one accountable team with humans in the loop.
02 — Capabilities
Everything a staff needs to be trusted with real work.
01 / Orchestration
Many agents, one staff
Assign roles, route tasks and let agents hand work to each other with the context intact.
02 / Shared memory
Knowledge that compounds
Decisions, preferences and outcomes persist across agents and sessions, so the staff gets smarter every week.
03 / Tool gateway
One governed connection
A single, permissioned gateway to email, CRM, documents and internal systems, scoped per agent and per task.
04 / Policies & approvals
Autonomy with limits
Define what agents may do alone and what needs a person. Approvals arrive with full context, in the channel you already use.
05 / Evaluation
Every run traced and scored
Replay any action, score outcomes against your own rubric, and catch drift before it costs you.
06 / Model-agnostic
Your models, your choice
Mix frontier models and open weights per task. Swap providers without rebuilding the team.
03 — How it works
Connect, define, run, review.
01
Connect
Link the tools your business runs on and the agents you already have. Each connection is scoped, logged and revocable.
02
Define roles
Describe the outcome, the boundaries and who approves what. SmartStaff turns it into a role any agent can take on.
03
Run
The staff works around the clock, sharing memory and handing tasks between agents. Anything outside policy pauses for a person.
04
Review and improve
Every run is traced and scored. Feed the results back, and the staff gets sharper with each cycle.
04 — Where it fits
Built for the work that never stops.
01 / Revenue operations
Pipeline that moves overnight
Research accounts, draft outreach in your voice, update the CRM and leave a review list for the morning.
02 / Customer support
Triage with context attached
Resolve the routine, escalate the rest with the full history, and never let a ticket go cold.
03 / Finance & back office
Reconcile, chase, file, report
The recurring work that keeps a business honest, done on schedule with every step logged.
04 / Engineering & product
Keep the codebase moving
Reproduce bugs, review pull requests, keep documentation current and flag what needs a human decision.
05 — Questions
Straight answers.
Does SmartStaff replace the agents we already use?
No. It sits above them. Keep your existing agents and frameworks; SmartStaff adds the shared memory, policy and oversight layer that makes them safe to run together.
Which models does it support?
Any model with an API, including the frontier providers and self-hosted open-weight models. Choose per task and swap at any time without rebuilding the team.
Where does our data live?
In your environment. SmartStaff can be deployed in your cloud or ours, in the UK or EU, and your data is never used to train third-party models.
How do approvals work?
You set policies for what agents may do unaided. Anything outside them pauses and asks a named person, with the full context, in the channel they already use.
How do we get started?
Request early access. We onboard a small number of teams each month and work alongside you for the first deployment.
Join the early access programme.
We are onboarding a small number of teams who want an AI staff they can trust with real work.