Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.cobalt.peoplereign.io/llms.txt

Use this file to discover all available pages before exploring further.

Capabilities overview

An agent’s capabilities are its tools. There are three ways to grant them, and they’re not interchangeable — there’s a clear order of preference.

Start with the highest level that works

1

Need to act in an external system? → Integration

Connecting your service desk, HRIS, identity provider, file store, or collaboration tool is almost always an Integration. Integrations are vendor-built, policy-gated, and audited — this is the primary path, and where you should look first.
2

No integration covers it? → Skillset

For a bespoke internal API or an unusual workflow that no integration handles, author a Skillset: natural-language Skills plus a Connection for credentials. This is the exception path.
3

Just reference material? → Custom knowledge

For documents that don’t live in a connected system — a policy PDF, an internal FAQ — add custom knowledge the agent can search and cite.

Why the order matters

IntegrationSkillsetCustom knowledge
Best forConnecting a known systemLogic an integration can’t doReference content
Built byCobalt (vendor-maintained)YouYou
MaintenanceLowHigher (you own the logic)Low
Acts in systemsYesYesNo (read-only)
Reaching for a Skillset when an integration exists means you take on maintenance Cobalt would otherwise carry, and you re-implement guardrails the integration already provides. Drop a level only when the level above genuinely can’t do the job.

Then control how tools behave

However a tool was granted, you constrain how it’s allowed to act with tools, rules & hooks: family rules, per-tool overrides, and hooks that gate individual calls. Privileged actions are gated and audited by default.