Solutions
Implementation Org Review Org Monitoring Managed Services
Agents
Discovery Agent Metadata Agent Design Agent Build Agent Test Agent Governance Agent Support Agent
Industry Solutions
Financial Services
Healthcare & Life Sciences
NDIS & Disability Services
Nonprofit
Not-for-Profit
Other Industries
Recruitment & Staffing Real Estate Cosmetic Procedures
Agentforce Claudeforce Blogs Pricing
Metadata Agent

Know your Salesforce org better than the people who built it.

One connection. The Metadata Agent reads every object, field, flow, class and permission, explains it in plain English, scores the health of the org, flags the risk — and keeps watching so nothing drifts back.

Get a free org health score See who it helps
Org Review / acme — production uat
Monitoring · delta refresh 14 min ago Run health check
Health score
0▲ 9
vs. 53 last month
Open findings
47▼ 18
3 critical · 11 high
Hours saved
126this month
vs. manual review & triage
Org health score
0 of 100
▲ 9 since last month
Security & permissions41
Technical debt55
Automation quality62
Data model68
Performance & limits74
Documentation38
Health score · last 6 months
acmepeer median
AprMayJunJulAugSep
Action plan · severity × effort
3 critical 11 high 33 medium
Critical
Modify All Data granted via 3 permission sets to 41 users
Security · PermissionSet · 41 users affected
2h Fix in Support
High
Recursive trigger on Opportunity — OpportunityTrigger ↔ Flow OppStageSync
Automation · Trigger + Flow · order of execution
1d Deep Research
High
27 hard-coded record IDs across 6 Apex classes
Technical debt · Apex · brittle across sandboxes
4h Create ticket
What changes when the org explains itself
70%
less time understanding a brownfield org
Every component explained in plain English before design starts — no more reverse-engineering weeks.
20+ hrs
saved per health check
A scored, prioritised assessment in about an hour, instead of two consultants for a week.
60%
faster impact analysis on change
Dependencies and downstream effects mapped before anyone touches production.
1 day
from connect to findings
OAuth into a sandbox on Monday morning; present the action plan to the client that afternoon.
What the Metadata Agent does

Four jobs. One agent. Zero manual org archaeology.

Understand, assess, monitor, and feed every other agent. Each pillar builds on the last, from the same single extraction.

01
Understand
Extract 60–70 metadata types and explain every file in technical and plain English. Fields with no description get one, inferred from how they are actually used.
Natural-language data dictionary
Auto-described fields, flows and classes
ER and process-flow diagrams on demand
Chat AI: ask the org anything
02
Assess
Score the org across twelve health lenses. Rank every finding by severity and effort. Deep Research any object into a five-minute best-practice report.
Org health score, benchmarked
Security, tech-debt and automation audits
Deep Research reports per component
Prioritised, developer-ready action plan
03
Monitor
Keep a delta-refreshed snapshot and watch for the signals that move an org toward risk. Material alerts become tickets automatically.
Scheduled or delta refresh
Drift, permission and failure monitors
Alert → Support Agent ticket, closed loop
Trend line for leadership
04
Feed every agent
The extraction becomes the shared context for Design, Build, Test, Governance and Support. Copy it across projects for the same customer — never re-extract.
Constrains Design Agent solutions
Grounds Support Agent triage
Baseline for Governance drift detection
Copy across projects, save tokens
See it in action

Connect. Ask. Watch. Fix.

Four screens from the Metadata Agent. Click a tab — each one plays through a real workflow.

BH
Brian Hughes
hughes.brian@company.com
Agents
Discovery Agent
Metadata Agent
Design Agent
Build Agent
Test Agent
Support Agent
Acme Corp — Production
Metadata listing · AI descriptions · process flows mapped
Extracted
ComponentAI descriptionUsage
Opportunity.Discount__cCustom field · Percent Approved discount applied at quote stage; referenced by 3 Flows and 1 validation rule. Active
Lead_AssignRecord-triggered Flow Routes inbound leads by region and product interest to the correct queue. Active
Account.Legacy_ID__cCustom field · Text Migration key from previous CRM; no references found in the last 12 months. Unused
OppAfterInsertApex Trigger Creates default line items and notifies the owner on new opportunities. Active
Case_EscalationProcess Builder Legacy escalation logic; overlaps with Case_SLA Flow. Overlap
Process flow · Lead → Opportunity
Lead created Lead_Assign (Flow) Qualified? Nurture (Trigger) Convert → Opp
2 Flows1 TriggerAuto-mapped
Chat with AIAcme Corp · Production
Explain the logic of AccountController.apex class and flag anything below 75% coverage.
9 Apex classes are below 75% coverage. Three of them touch the Patient__c trigger path — that’s where I’d start.
ApexEncounterReassignment.cls62% · 318 lines
ApexPatientConsentBatch.cls64% · 184 lines
ApexAccountController.cls71% · 402 lines
Ask anything about your org…Deep ResearchGenerate Assets
Workspace · Insights 3
EncounterReassignment — coverage watch
Every 15 min · 1 fire
Deep Analysis: Account object
Saved 2 Jan · Process Insights
Governance & hygiene posture
Saved 2 Jan · Process Insights
On match →
Create ticket in Support AgentTrigger
Monitors
Continuous agents watching Acme Corp · Production
+ New monitor
Critical
Modify All Data on non-admin profiles
All profiles + perm sets · Every 15 min
1 finding · ticket T-1827
High
Sharing rule on Patient__c opened wider
Patient__c · Sharing rules · Real-time
No findings
High
Apex coverage below 80%
Org-wide · On every deploy
1 finding · ticket T-1825
Medium
Custom field added on Patient_Encounter__c
Patient_Encounter__c · Every 15 min
1 finding
Low
License utilisation > 90%
Salesforce CRM license · Daily 06:00
Paused
CRITICALT-1827 · auto-created by monitor Modify All Data on non-admin profiles · 3 min ago
Integration User profile gained Modify All Data on Patient__c
Impact analysis · from metadata
Profile Integration_User is assigned to 41 active users, 3 of them external.
Patient__c holds 12,480 records including PHI fields (DOB, Medicare_No__c).
2 flows and 1 trigger write to Patient__c under this profile.
No change request found in the last 30 days for this permission.
Proposed fix
Remove ModifyAllData from Integration_User; grant object-level CRUD on Patient__c only. Deploy via Build Agent to UAT first.
Detected by monitorMetadata Agent
Impact analysedMetadata Agent
Fix proposedSupport Agent
Deploy to UATBuild Agent
Org health check

Twelve lenses. One score. A roadmap you can hand to a developer.

The health check runs against the extracted metadata — not a checklist. Every finding is ranked by severity and effort, explained in plain English, and linked back to the exact component so remediation starts immediately.

Architecture & Well-Architected
Coupling, boundaries, scalability, resilience.
Technical debt & complexity
Obsolete config, duplicates, brittle customisation.
Security, privacy & compliance
Least-privilege gaps, sharing risk, dormant access.
Automation quality
Recursion, overlap, fault handling, order of execution.
Data quality & model health
Completeness, duplication, unused fields, skew.
Performance & scale
Slow transactions, selectivity, limits pressure.
Usage & adoption
Dead reports, licence utilisation, process bypass.
Release health & drift
Prod vs sandbox differences, unmanaged change.
Integration health
Auth, endpoints, failure handling, observability.
Maintainability & operability
Documentation, ownership, testability, admin effort.
Business-process fit
Config that no longer matches how the business works.
AI & Agentforce readiness
Data trust, permission boundaries, grounding quality.
Deep Research on any component

Pick a custom object or ask a question. The agent drafts a research plan — summary, relationships, automation, security, best-practice gaps — you approve it, and a full report lands in about five minutes. Save it as an Insight and it becomes part of the org’s permanent knowledge base.

20+hrs
of manual health-check effort replaced per org, per review. Schedule it monthly and the score becomes a trend line you can show leadership.
Continuous org monitoring

A health check is a photo. Monitoring is the film.

Orgs don’t stand still between reviews. New releases, permission changes, failing integrations and creeping automation quietly move a healthy org toward risk. The Metadata Agent keeps a delta-refreshed snapshot and watches for the signals that matter.

Permission & sharing drift
New Modify All / View All grants, public groups widened, dormant admins.
Automation failures
Flow fault rate rising, batch jobs failing, triggers throwing limits.
Release & environment drift
Production diverging from source control or from the last approved snapshot.
Integration health
Endpoints erroring, auth expiring, callout volume spiking.
Complexity creep
Field and flow counts climbing on core objects without a matching design change.
Explore Org Monitoring
Alert stream · liveWatching
Critical
Modify All Data granted to Integration User profile
Detected 3 min ago · PermissionSet · Security
Ticket SUP-4193
High
Flow failure rate on Order Fulfilment up 340% in 24h
Detected 18 min ago · Flow · Automation
Ticket SUP-4192
Medium
12 new custom fields on Account with no description
Delta refresh · Data model · Documentation
Auto-describe
Low
Sandbox UAT drifted from production — 6 components
Nightly compare · Release health
Review diff
Every material alert → ticket in Support AgentClosed loop
Who it helps

Built for the people who inherit the org.

Five roles, one extraction. Pick a seat.

Solution Architect
70%
less time reverse-engineering before design
Design against what actually exists — not what the last consultant remembered.
Plain-English dictionary of every object, field, flow and class
ER and process-flow diagrams on request
Deep Research on relationships, automation and best-practice gaps
How it works

Connect on Monday. Present the findings on Tuesday.

OAuth into any environment. The agent handles extraction, explanation and scoring — then keeps the snapshot current with delta refreshes you can schedule.

1
Connect
OAuth into sandbox, production or a custom domain. No password stored.
2
Extract
60–70 metadata types pulled in 5–10 minutes. Metadata only — no records.
3
Explain & score
Every file described; twelve lenses scored; findings ranked by severity and effort.
4
Act
Deep Research, Chat AI, diagrams. Findings become tickets in the Support Agent.
5
Monitor
Delta or scheduled refresh; alerts on drift, permissions and failures.
The context engine

One extraction powers every agent that follows.

Connected metadata is copied — not re-extracted — across projects for the same customer. It constrains the Design Agent, grounds the Support Agent and gives the Governance Agent a baseline to detect drift against.

Design Agent
Uses the extraction as a hard constraint when writing low-level solutions — no designing against fields that don’t exist.
Build Agent
Generates configuration and code that fits existing patterns, naming and automation order.
Support Agent
Identifies dependencies and impact for every ticket from the same connected metadata.
Governance Agent
Detects scope gaps and environment drift against the extracted baseline.
Free org health score

See your org the way the agent sees it — in one working day.

Connect a sandbox, get a scored health check across twelve lenses, and walk away with a prioritised remediation plan. No obligation, no record data touched.

Book my free health score
What you get
Full org extraction and plain-English dictionary
Health score across twelve lenses, benchmarked
Prioritised action plan ranked by severity × effort
Deep Research report on your most complex object
Connected context ready for every other agent