Capture
The Recorder
Do the work once in Chrome — it becomes a step-by-step guide anyone can follow.
Your org changes every day. Your agents keep up.
AI agents for yourSalesforce team.
A crew of agents on your org — answering questions, guiding users live, sweeping your metadata every 15 minutes, writing it all down. They do the work; you review it.
No managed packageMetadata only — never your recordsHuman approval on every write
Reads from · writes to · renders
SalesforceGitHubGitLabConfluenceChromeSlackClaudeCursorClineMCPMarkdownOpenAPIMermaidD2draw.ioExcalidrawSalesforceGitHubGitLabConfluenceChromeSlackClaudeCursorClineMCPMarkdownOpenAPIMermaidD2draw.ioExcalidraw
The problem
Your org changes faster than any human can keep up.
Flows, fields, approval steps, permission sets — something moves every week, and you can't hire your way out of it. So everyone keeps working from yesterday's truth:
- New admins follow steps that no longer exist
- Senior people answer the same Slack question, again
- Your AI tools confidently quote the stale page
The crew
Six agents.
One for every job in your org.
Admins, end users, RevOps, even your AI tools — each has an agent. Every one proposes; you approve.
Guide Me
The Guide
Walks any user through Salesforce live, spotlight on the exact next button.
Drift & commits
The Watcher
Sweeps your metadata and tracked branches every 15 minutes, then drafts the fix.
Coverage
The Auditor
Finds what's undocumented and files ready-to-ship drafts.
Doc copilot
The Editor
Inline suggestions in the editor — accept or reject hunk by hunk.
Org intelligence
The Analyst
Impact analysis, who-can-access, org health — read-only, straight from your metadata.
Org intelligence
Your org, explained.
Not just documented.
Ask in plain English. The agent loops read-only tools over your synced catalog and your docs — and shows its work.
IMPACT
What breaks if I change this?
Every usage of a field, class or flow — and what it depends on before you deploy.
ACCESS
Who can see or edit this?
Access resolved through profiles and permission sets, plus a live security check.
HEALTH
How healthy is the org?
A 0–100 score with trend, unused fields, count-only utilization. Never record contents.
HISTORY
What changed last week?
Before/after diffs for everything that moved, plus the live Setup Audit Trail.
SCHEMA
Show me the data model.
An interactive ER diagram built from synced metadata. One click to the docs.
ANALYTICS
Chart won deals by month.
Ask for it; the chart renders natively over your own BI instance. Read-only.
Or run a playbook
/audit/changes/stale/coverage/readiness/classify
Ask the agents
Ask anything about your org. Get a grounded answer.
Ask in the app, in the Chrome extension, or from Claude Code, Cursor and Cline over MCP. Answers cite the doc and the metadata behind them. The same tools can write — always as a change request, never a silent edit.
Ask in the appAsk-docs in ChromeMCP for Claude, Cursor & Clinelc CLI
> how do opps over $250k get approved?
● livingcontext-mcp · search_docs · get_metadata_context
→ opp-approval.md · cites ApprovalProcess v3.1
Routes to Regional-VP, two-of-three quorum,
requires ContractTerm__c since v3.1.
✎ update_doc → change request opened for review
How it works
The agents do the work. You approve it.
01
Put the crew on your org
Connect GitHub or GitLab — your docs live there and stay there — then point us at Salesforce. The catalog syncs on a cadence you pick.
02
They do the work
A sweep every 15 minutes: new commits, metadata that moved, components nobody documented. Plus questions answered and users guided live, all day.
03
You keep the final say
Drafts land in an inbox you triage one at a time. Accept, and it becomes a change request or a PR in your own repo — a real diff, edit by edit.
0
specialist agents working your org, around the clock
0%
of agent writes gated behind human approval — on every plan
0
read-only tools the crew calls across your org and docs
47
questions your admins have to answer twice
The paper trail
Everything they do, written down.
Documentation isn't the product — it's what falls out of the agents doing the work. It lands in a fast, searchable site backed by a repo you own.
- A real docs site — tabs, sections, and a changelog page
- Full-text search across everything (⌘K)
- API reference generated from your org, with a live “Try it”
- Diagrams the agents author and lint: Mermaid, D2, draw.io, Excalidraw
- Templates plus a house style every draft is written to
- Real-time co-editing, with tracked changes you accept per hunk
- Confluence spaces imported as MDX, opened as a pull request
- Org subdomains, spaces, and role-based permissions
- Git-backed: history, blame and rollback for every word
Before / after
Life before the crew vs. life after.
Your team today
- Screenshots from 2023, still in the onboarding doc
- The runbook says Sales-Director; the org says Regional-VP
- Every new admin asks the same questions in Slack
- Your AI assistant confidently quotes the stale wiki
- Nobody knows which pages are still true
Your team with the crew
- Guides recorded in one take, screenshots always current
- Metadata and commits swept every 15 minutes — fixes drafted
- New admins follow a spotlight inside Salesforce
- Assistants answer from your docs and live metadata, with citations
- Coverage and freshness, visible on one dashboard
Is this just a documentation tool?
No. It's a crew of agents that work your org — answering questions, tracing impact and access, guiding users live, watching what changes. Documentation is the paper trail they leave behind: current, git-backed, yours.
Who is this actually for?
Everyone who touches your Salesforce. Admins capture processes, end users get live walkthroughs and plain-English answers, RevOps see coverage and drift on one dashboard, and your AI tools read grounded context over MCP.
Do I need to install anything in Salesforce?
No managed package. Capture and Guide Me run from a Chrome extension; sync uses Salesforce's standard APIs over OAuth — enough to read your org's shape, nothing more. Production and sandbox connect side by side.
Where do my docs actually live?
In a git repository you own, on GitHub or GitLab. The docs site renders straight from it, so every page has history, diff, blame and rollback — and it's all portable if you ever leave.
What does the AI get to change on its own?
Nothing. Doc work lands in an inbox, a change request or a PR — you approve per hunk, not per document. Anything with a side effect outside LivingContext parks as a pending approval; no plan removes that gate, and everything that runs hits an immutable audit log.
What about sensitive data in recordings?
Redaction happens in the browser, at capture time. Passwords are always masked — that floor can't be turned off — smart blur runs as you record, capture is limited to the browser tab, and nothing leaves until you publish.
Can my AI tools read the docs?
Yes — and write to them. Every org gets an MCP server: docs reads, Salesforce org intelligence, reviewed writes. Point Claude Code, Cursor or Cline at it, or use the lc CLI. Keys are org-scoped, membership is re-checked every call, and writes land as change requests.
How fast do the agents notice a change?
A shared sweep runs every 15 minutes — new commits, coverage gaps that are due, sync schedules that have come around. Catalog cadence is per connection: manual, hourly, 6-hourly or daily. Anything can also be run on demand.
Will this survive our security review?
It's built for one: two-factor, enterprise SSO (Okta, Entra, Auth0), tokens encrypted at rest, single-view API keys that die when a member leaves, an org IP allowlist, an append-only audit log with your retention window, and a one-switch opt-out that stops all content going to AI providers. SCIM is on the roadmap, not shipped — we'll say so up front.
Closed beta · Q3 2026 · limited seats
Six new teammates.
Zero new headcount.
Your org will change again tomorrow morning. Put the crew on it — answering, guiding, watching and writing it all down — while you just review their work.