Reference

Capabilities.

The complete list Nine groups Updated 7 September 2026

This page doesn’t argue anything. It’s the full set of what the platform does, grouped and stated plainly, for anyone evaluating rather than being sold to. If you want the case for why, that’s how it works.

01Agents and roles 02Skills and knowledge 03Memory 04Permissions, approvals and audit 05Scheduling and autonomy 06Workflows and outputs 07Connectivity 08Models 09Deployment, brand and administration

01 · Agents and roles

Agents and roles

RefCapabilityState

1.1Agents as processes. Each agent runs in its own sandboxed environment with its own identity, memory and permissions, rather than as a session inside a chat window.Live

1.2Out of the box agents. Pre-built agents for recurring operational roles, ready to deploy against your systems.Live

1.3Custom agents. Defined with you during onboarding, then built and trained for work that’s specific to your business.Live

1.4Self service creation. Your own team can create agents and skills inside the platform, within guardrails you set, without engineering.Live

1.5Templates. New agents start from a template carrying a system prompt, sensible defaults, starter skills and recommended connectors.Live

1.6Teams. Agents can be grouped into teams that map to how the business is organised.Live

1.7Per agent tuning. Tone, verbosity and reasoning tier are set individually rather than globally.Live

02 · Skills and knowledge

Skills and knowledge

RefCapabilityState

2.1Skills as written procedures. Plain text handbooks describing how a job is done, readable and editable by anyone on your team.Live

2.2Shared library. Skills are held centrally and distributed to every agent that needs them, rather than copied per agent.Live

2.3Bundled and custom. A catalogue of pre-written skills sits alongside your own, and yours override a bundled skill of the same name.Live

2.4Self authored skills. An agent can write a skill for itself as it learns a repeatable procedure, and those are always on.Live

2.5Trigger definitions. Each skill states when it should be used, so work routes to the right procedure without being told.Live

2.6Versioned and portable. Skills are files. They can be inspected, changed, tracked and exported.Live

03 · Memory

Memory

RefCapabilityState

3.1Organisational memory. What agents learn about the business is held by the business, not inside one person's history.Live

3.2Readable and editable. Memory is plain text your team can open and correct when it’s wrong.Live

3.3Overnight consolidation. A scheduled pass re-reads the day, promotes what’s durable, retires what’s stale and files its own follow ups.Live

3.4Survives a model change. Memory sits above the model, so swapping the model underneath loses nothing.Live

3.5Topic organisation and search. Entries are categorised and searchable rather than an undifferentiated log.Live

3.6Exportable. Yours to take, readable without our software.Live

04 · Permissions, approvals and audit

Permissions, approvals and audit

RefCapabilityState

4.1Allow, ask, deny. Three stances set per individual tool, not per application.Live

4.2Scoped by person, agent and tool. The same agent can behave differently depending on who is asking.Live

4.3Value thresholds. Act freely below a limit, ask above it. Set per agent and per action.Live

4.4Credentials never reach the model. Held in a dedicated encrypted store, resolved at call time, redacted on the way back. A prompt injection can raise a request, never take an action.Always

4.5Approvals as durable state. Requests persist across restarts, expire on a known clock and are recorded with their outcome.Live

4.6Approvals where your team works. Requests arrive in the channel already in use rather than in a separate console.Live

4.7Full audit trail. Who asked, which agent, which tool, which credential, the policy decision, the approver and the result.Live

4.8Exportable to your pipeline. Streams to S3, Datadog, Splunk or your own SIEM. Retention follows your policy.Live

4.9Role based access for people. Administrators control what each person can see, reach and change.Live

05 · Scheduling and autonomy

Scheduling and autonomy

RefCapabilityState

5.1Scheduled jobs. Recurring work on a defined cadence, delivered to a channel, with run history and transcripts.Live

5.2Event driven triggers. Work starting from something happening in a connected system rather than from a clock.Live

5.3Self scheduling. An agent can schedule itself for later, in minutes, tomorrow or every Monday, so work spanning days survives between sessions.Live

5.4Multi day tasks. A thread opened on Monday can be picked up on Friday by the agent that opened it.Live

5.5Run history. Every execution recorded with timestamp, outcome and a readable transcript.Live

5.6Unattended operation. Work proceeds with nobody present, pausing only where a stance requires a person.Live

06 · Workflows and outputs

Workflows and outputs

RefCapabilityState

6.1Shared boards. Agents and people own different columns of the same board, with work moving between them.Live

6.2Customer defined stages. Columns match how your business already works rather than a process we impose.Live

6.3Typed transitions. Each move carries a rule. Some fire automatically and wake the next agent, others hold for a person.Live

6.4Dashboards. An agent composes a live reporting surface from the data it can reach, updated on a schedule and shareable.Live

6.5Applications. Working tools and interfaces built on your connected data, launched from the workspace.Live

6.6Documents and artefacts. Reports, packs and files produced as attachments rather than pasted into a chat.Live

07 · Connectivity

Connectivity

RefCapabilityState

7.1100 plus connectors live. Across commerce, marketing, support, finance, analytics, inventory and communication.Live

7.2Native packages. Hand tuned integrations for the systems customers live in, with curated actions rather than raw API surface.Live

7.3Open standard. Built on the Model Context Protocol, so the library grows through the wider ecosystem as well as our releases.Live

7.4Your own APIs. Internal systems, warehouses, partner endpoints and bespoke services lifted in as tools.Live

7.5Central connector catalogue. Connect once, allocate to the agents that need it, revoke in one place.Live

7.6Standard authentication. Connections made through normal sign in flows, with credentials encrypted at rest.Live

7.7Built on request. Connectors that don’t exist yet are defined with you during onboarding, and adding one is a configuration exercise rather than a rebuild.Live

08 · Models

Models

RefCapabilityState

8.1Model agnostic. Not dependent on any specific LLM. Models are configurable by design.Live

8.2Nothing you build lives in the model. Agents, skills, memory, connections, automations and applications sit on the platform, so changing the model leaves all of them untouched.Live

8.3Failover to a backup model. If the default model has an outage, work fails over to a configured backup, so agents keep running without downtime.Live

8.4Reasoning tiers. Lighter and cheaper, or stronger and slower, chosen per agent rather than for the whole account.Live

8.5No training on your data. Providers are configured so your content isn’t used to train their models, and we don’t train our own on it.Always

09 · Deployment, brand and administration

Deployment, brand and administration

RefCapabilityState

9.1Three deployment options. Our managed cloud in Sydney, entirely inside your own AWS, GCP or Azure account, or hybrid.Live

9.2Runs without us. The runtime is self contained with local configuration and local credential storage, and doesn’t phone home.Live

9.3Tenant isolation. One workspace per customer with its own storage. Real separation, not a shared database with a partition.Live

9.4White label. Your logo, accent colour, typeface and display name across the whole interface.Live

9.5Custom domain. Runs on your own subdomain rather than ours.Live

9.6Unlimited people. No per seat fee on any plan, with access granted per person and per agent.Live

9.7Usage visibility. A live meter showing what the workforce is costing, with a cap you set.Live

9.8Single sign on. Available on the Scale plan.Scale

9.9Full export. Agents, skills, memory and configuration. Off boarding is supported rather than obstructed.Live

Next step

Checking whether we do something specific?

If it isn’t on this list, ask. Half of what’s here was built because a customer needed it and told us. Thirty minutes on a call and we’ll show you the relevant part running rather than describing it.

Book a demo → See it in context →