MCP server and integrations

Works wherever your team already works.

Our MCP server connects Claude Code, Cursor, Copilot or Agentforce, or your agent of choice, to the metadata graph database. Everything else here is another way into Elements, because you should not have to move to get an answer about your own Org.

Salesforce MCP server for AI coding tools and agents

Elements exposes an MCP server, so any client that speaks the protocol can reach your Org model. These four are the ones we build and test against.

Anthropic

Claude Code

The fullest implementation. Ships as a plugin rather than a bare server, so the MCP tools arrive bundled with the skills and sub-agents that know how to drive them. Long jobs run in phases with state on disk.

Available

Anysphere

Cursor

Connect the MCP server and ask about your Org without leaving the editor. Same graph database, same resolved dependencies, alongside the code you are already writing.

Available

Microsoft

Copilot

The same server, reached from Copilot, so teams standardised on Microsoft tooling get the Org context without adopting anything new.

Available

Salesforce

Agentforce

Ground Salesforce's own agents on the resolved model rather than on raw metadata, so their answers are based on your actual Org rather than on a guess at it.

Available

Any other MCP-compatible client can connect too. These four are the ones we build and test against.

A server is not the same as a plugin.

Most MCP servers hand an agent a set of tools and leave it to work out the rest. A tool tells an agent what it can do. It does not tell it when, in what order, or what good looks like.

Tools

The calls that read your Org. Every client gets these.

Skills

Which tool to reach for and in what order. Bundled where the client supports it.

Phases

State written to disk between phases, so a six-hour job survives a restart.

The technical detail, if you want it.

Seventeen skills, nine engagements and twenty MCP tools. Two grids showing exactly how tools, skills and phases compose. Reference rather than reading, so hover a row or a column.

Which skills go into which engagement

Seventeen skills across nine engagements. Hover any row or column to see what connects to what.

Which of the seventeen skills each of the nine Salesforce engagements uses
Skill Engagement	Org discovery	Agentic roadmap	Impact assessment	Org merge / split	CPQ → Revenue Cloud	Data Cloud set-up	Data corruption	Access audit	Clean up & agentify
Live skills
Metadata ExplorerFinds and explains individual metadata nodes: objects, fields, flows, triggers, validation rules.	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Org DiagnosticBaseline health check across structure, automation, tech debt, compliance and governance.	Yes	Yes	No	Yes	No	Yes	Yes	No	Yes
Tech DebtResolves flagged nodes, checks blast radius and last-touched date, returns a ranked remediation list.	Yes	Yes	No	Yes	Yes	No	Yes	No	Yes
Access AuditWhich Profiles, Permission Sets and Groups grant access to a node, and how many users that is.	Yes	No	No	Yes	No	Yes	No	Yes	No
Change BriefingEvery node changed in a date window, with velocity, breakdown by type and anomalies.	Yes	No	Yes	No	No	No	Yes	Yes	No
Metadata QueryFiltered ad-hoc lists of metadata, saveable as a named custom view.	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Package UninstallIdentifies every dependency that would break, and sequences the remediation first.	No	No	No	Yes	No	No	No	No	Yes
MetaFieldsCreate, update and query custom metadata fields, layering your data onto the metadata graph database.	Yes	No	No	Yes	Yes	Yes	No	Yes	Yes
Agent OpportunitiesScans for automation candidates and scores each by effort and impact.	No	Yes	No	No	Yes	No	No	No	Yes
Org-to-ProcessReverse-engineers the business process hidden inside existing automation.	Yes	Yes	No	Yes	Yes	No	Yes	No	Yes
Agent FinderScores every process against an agent-readiness rubric.	No	Yes	No	No	Yes	No	No	No	Yes
AgentScriptDrafts and previews Agentforce topic specs before anything is committed.	No	Yes	No	No	Yes	No	No	No	Yes
DiagramsAuto-generates visual process maps from live configuration.	Yes	Yes	No	Yes	Yes	Yes	No	No	Yes
StoriesTurns a process diagram into a backlog with acceptance criteria, ready for Jira or ADO.	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
In development
Agent Diagnostics ↗In development. Monitors deployed agents for misfires, escalations and config gaps.	No	Yes, in development	No	No	No	No	Yes, in development	No	Yes, in development
Decision Engine ↗In development. Models the full blast radius of a change before it reaches production.	Yes, in development	Yes, in development	Yes, in development	Yes, in development	Yes, in development	Yes, in development	Yes, in development	No	Yes, in development
Org-to-Agent ↗In development. Turns a recovered process map into a complete Agentforce agent design.	No	Yes, in development	No	No	Yes, in development	No	No	No	Yes, in development

LiveIn developmentHover a row for what the skill does, hover a cell to trace the column.

And which MCP tools each skill orchestrates

The layer underneath. A skill is not one call, it is a sequence of them in the order our architects run them.

Which MCP tools each skill orchestrates
MCP tool Skill	Metadata Explorer	Org Diagnostic	Change Briefing	Tech Debt	Access Audit	Metadata Query	Package Uninstall	MetaFields	Agent Opportunities	Org-to-Process + AgentScript	Diagrams	Stories	Org-to-Agent ↗	Agent Diagnostics ↗	Decision Engine ↗
list_reference_models	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes	Yes	No	No	Yes	No	No
metadata_search	Yes	No	No	No	Yes	No	Yes	No	No	Yes	No	No	No	No	Yes
get_metadata_node	Yes	No	Yes	No	Yes	No	Yes	No	No	No	No	No	No	No	Yes
get_dependencies	Yes	No	No	Yes	No	No	Yes	No	No	No	No	No	No	No	No
explain_metadata_item	Yes	No	No	No	No	No	No	No	No	No	No	No	No	Yes	No
get_org_structure	No	Yes	No	No	No	No	No	No	No	No	No	No	No	No	No
get_automation_health	No	Yes	No	No	No	No	No	No	No	No	No	No	No	No	No
get_tech_debt_summary	No	Yes	No	Yes	No	No	No	No	No	No	No	No	No	No	No
get_analytic_drill_down	No	Yes	Yes	Yes	No	No	No	No	No	No	No	No	No	No	No
get_node_change_history	No	No	Yes	Yes	No	No	No	No	No	No	No	No	No	No	No
query_metadata	No	No	No	No	No	Yes	Yes	Yes	Yes	No	No	No	No	Yes	No
list_metafields	No	No	No	No	No	No	No	Yes	No	No	No	No	No	No	No
scan_agentic_opportunities	No	No	No	No	No	No	No	No	Yes	No	No	No	No	No	No
generate_org_to_process	No	No	No	No	No	No	No	No	No	Yes	No	No	Yes	No	No
get_process_map	No	No	No	No	No	No	No	No	No	Yes	Yes	No	Yes	No	No
preview_agent_script	No	No	No	No	No	No	No	No	No	Yes	No	No	Yes	Yes	No
generate_process_map	No	No	No	No	No	No	No	No	No	No	Yes	No	No	No	No
generate_stories_from_diagram	No	No	No	No	No	No	No	No	No	No	No	Yes	No	No	No
run_change_impact_analysis ↗	No	No	No	No	No	No	No	No	No	No	No	No	No	No	Yes, in development
run_backlog_draft ↗	No	No	No	No	No	No	No	No	No	No	No	No	No	Yes, in development	Yes, in development

LiveIn developmentEighteen live tools and two in development. One skill is many calls in a known order.

Every row on the left is answered in the Elements app in a few clicks, or by asking, if your team works that way. Either route, you stopped doing the middle column.

Book a demo

The connection everything else runs on

The MCP server, the app, the Chrome extension and every tool above are ways of reaching something Elements already knows. That knowledge comes from one place: your Salesforce Org.

Every component, indexed

Objects, fields, flows, Apex, validation rules, layouts, permission sets, reports and list views. Read from the Org and held in the metadata graph database, with the relationships resolved rather than inferred.

Production and sandboxes

Connect as many Orgs as you run. Pricing is on your total Salesforce Org user footprint rather than per connection, so a multi-Org estate does not get penalised for being one.

Kept current as the Org changes

Re-indexed on a schedule with a change log, so the picture stays current instead of ageing from the day someone last wrote it down.

Beyond the native dependency API

Salesforce exposes MetadataComponentDependency, but it is capped at 2,000 records per query and you cannot report on it. Elements resolves multi-level dependencies across the whole Org.

Industry Clouds

Industry Cloud metadata is indexed automatically with the Industries add-on, so OmniStudio and the rest of it stops being a blind spot.

Data Cloud

Data 360 metadata is indexed automatically with the Data360 add-on, and appears in the same graph as everything else.

Delivery and change tooling

Requirements and stories generated from the real Org, pushed where your team already tracks work, with the metadata context attached.

issue tracker

Jira

Stories generated from the real Org, pushed into your board with the dependency context attached.

issue tracker

Azure DevOps

The same story output, into Azure Boards.

pipeline

Copado

Analysis handed to the pipeline your team already deploys through.

and the rest

All DevOps integrations

Including Salesforce DevOps Center. The full list lives on its own page.

More ›

And everywhere else you already are

The same graph database, reached without leaving the tool you happen to be in.

in the org

Managed package

Installed from AppExchange, so the dictionary and impact analysis sit inside Salesforce.

in setup

Chrome extension

Metadata context on the Setup screen you are already looking at.

More ›

standalone

Metadata Explorer

The metadata explorer on its own, outside the full space.

More ›

Not sure which route fits your team?

Tell us what your team works in and we will show you what connecting looks like.

Book a demo✦✦Read the docs