Agents overview

One engine, many roles — interactive agents you chat with and autonomous roles that keep documentation fresh, all shipping through human review.

LivingContext is built on a single agent engine that takes on different roles. Some roles are interactive — you chat with them directly. Others are autonomous — they watch your Salesforce org and code repositories, then draft documentation changes on their own. Every role shares the same grounding: your synced Salesforce catalog, your docs corpus, and your connected repositories.

One engine, many roles

Interactive agents

Workspace agent

Ask anything about your Salesforce org and docs from Home. Read-only org intelligence with live tool activity and slash playbooks like /audit.

Doc editor agent

Chat about the doc you’re editing in a change request. Proposes surgical edits as inline tracked changes you accept or reject.

Autonomous roles

Commit updater

Watches the branches you track and drafts documentation updates for each commit that lands.

Coverage auditor

Finds important Salesforce components that have no documentation at all and proposes brand-new docs for them.

Want to go beyond the built-in roles? The Agent platform lets you define custom agents with their own personas, tools, triggers, and approval rules.

Grounded or it doesn’t ship

No agent in LivingContext publishes anything directly. Every proposal — whether it’s an updated paragraph, a new page, or a saved answer — travels through a human review surface before it reaches your documentation:

Surface	What you review there
Inbox	Drafted doc updates and new-doc proposals, triaged one at a time
Change requests	Full editing sessions with diffs, comments, and tracked changes
Pull requests	The final merge into your Git repository — your repo, your history

Agents must also show their work. Drafts are grounded in real evidence — actual Salesforce metadata, actual commits, actual source — and proposals without evidence are skipped rather than guessed.

Because your docs live in your own GitHub or GitLab repository, every agent-authored change ends up as an ordinary commit you can inspect, revert, or audit later.

How often agents run

Autonomous roles run on a shared scheduler that sweeps every 15 minutes. Within a sweep, each role checks whether it has work to do:

• The commit updater walks tracked branches forward through new commits, a few per sweep, so a burst of commits drains steadily rather than flooding your Inbox. Newly tracked branches start from their current tip — no backlog dump.
• The coverage auditor checks whether its configured frequency (daily, weekly, or monthly) is due.
• Salesforce metadata sync schedules fire according to their own cadence.

You can always trigger work manually — for example, the coverage auditor has a Run now button, and custom agents on the Agent platform support manual triggers.

Where to go next

Workspace agent

The full tool catalog and slash playbooks.

Doc editor agent

Tracked-change editing, mentions, and AI skills.

Coverage auditor

Configuration and how proposals reach the Inbox.

Agent platform

Build, trigger, and govern your own agents.