Analyze, for Salesforce orgs

Your Salesforce org, read from its source.

Upload the metadata you already pull and ask in plain language. Get answers cited to the file and line and to Salesforce's own documentation, diagrams you edit, and models you run again with your own numbers. Read-only by design.

Analyze a workspace Salesforce FAQ

A Salesforce Flow diagram in Symposium with record-changing elements highlighted and an inspector warning about an update inside a loop with no fault path — A cancellation Flow drawn like Flow Builder. Highlight DML marks every record change; the inspector flags an update inside a loop with no fault path.

Never connects to your org: You upload metadata you already pulled. No org credentials, no deploys, no data writes.
Your metadata, your sandbox: The workspace is read in a private sandbox for your session, with no internet access.
Org facts from your own CLI: When metadata isn't enough, your Claude Code or Codex gathers the facts read-only.

01Twelve jobs

Start from what you want to come away with.

Pick the workspace, then the job: an answer, a diagram, a model or a report. Or ask your own question and let the AI choose.

The Analyze a Salesforce workspace dialog, showing job choices such as Explain how something works, Draw the data model and Who can see these fields — The workspace dialog. Every job says what it gives back and what you can do with it afterward.

Answer

Read in the chat.

Explain how something works
A Flow, a field, a trigger or a process, explained from its source.

Diagrams

Edit by hand or by asking.

Draw the data model
Objects, fields and how they relate.
Show a Flow and what it starts
A Flow drawn like Flow Builder, with the flows it starts.
What runs when a record saves
Every automation on an object, in the order Salesforce runs it.

Models

Change the inputs and run them again.

What a change would touch
Rename, delete or retype a field and see what refers to it.
Who can see these fields
Which profiles and permission sets can read or edit them.
When we'll hit an org limit
Storage, email or API limits forecast from your org's readings.

Reports

Publish as PDF, PowerPoint, Word or an interactive page.

Troubleshoot an error
Paste a Flow error, an Apex exception or a bug report.
Audit the workspace
Risks, conflicts and what to fix first.
Review access and governance
Profiles, permission sets, sharing and broad permissions.
Get ready for a release or a feature
What a Salesforce release or a planned feature touches, and a checklist.
Hand off to a developer
Acceptance criteria and implementation prompts, never patches.

02Diagrams

Your org, drawn the way you work on it.

Diagrams come straight from your metadata. Move the boxes, add fields or steps, or ask the AI to make a change. Every save is a version you can go back to.

An editable data model diagram of Contact, Booking, Experience, Session, Credit, Guest Review and Guest Incident objects with their fields — The data model. Your own objects first, each with its sharing model and record types. Objects from installed packages appear only where yours point at them, collapsed under their namespace.

A save-order diagram for Booking__c: system validation, a before-save Flow, an Apex trigger, then validation rules, with the selected Flow described in a side panel — What runs when a record saves. Every automation on an object in the order Salesforce runs it: before-save Flows, triggers, validation rules, after-save Flows and roll-ups, each tied to its file.

symposiumai.app/analyze

Flows that read like Flow Builder. Turn on Highlight DML and every record change lights up. A record operation inside a loop is flagged, and so is a step with no fault path.

03Models

Models you run again without the AI.

A model has inputs you change and results that update when you run it. Name a scenario, compare any two, or download the model to run on your own computer.

symposiumai.app/analyze

What a change would touch. Rename a field and see every Apex class, Flow, formula field, validation rule, layout, Lightning page, report and email template that refers to it, each with its file and line.

A data storage forecast model: 3 of 5 MB in use, growth set by hand at 10% a month, reaching 80% and 100% of the limit on given dates — When we'll hit an org limit. A forecast from the dated readings in your evidence packets. Fit the growth to the readings or set it yourself; the result says how many readings, over how many days, it was fitted to.

04Answers you can check

It cites what it read.

Answers and reports separate what your metadata shows, what the org showed, what the documentation says, and what is still an assumption.

Source claims cite the file and line: When the AI says what your org does, it points at the metadata that does it: a Flow element, a line of Apex, a validation rule.
Platform claims cite the documentation it read: A limit, what a setting does, the order of execution: each cites a passage the AI received from official Salesforce documentation. If no passage says it, the claim is marked unverified.
A report can't cite what the AI never read: A report that presents documentation the AI never received is refused until the AI cites the passage or marks the claim unverified.
The Reviewer checks each claim: A second AI checks every Salesforce claim in a report against the passage it cites. Check independently has a different AI redo any point.

An Analyze answer explaining what runs when a booking is canceled, with inline citations to Flow files and line numbers — An answer that walks through a cancellation in the order Salesforce runs it, citing each Flow and rule by file and line.

A report's Sources panel showing a Salesforce Help article, its retrieval time and the passages the AI read — A troubleshooting report's Sources: the Salesforce Help page the AI read, when it read it, and the passages it received.

05Troubleshoot

Paste the error. Get the cause.

Paste a Flow error, an Apex exception or a bug report. The AI traces it through your metadata to the step that fails and the rule behind it, then writes the fix as tasks, not patches.

A troubleshooting report's diagnosis: the root cause in a cancellation subflow, the failing step, the credit formula and when it fails — A cancellation fails with “Credits over $500 need a reason.” The report finds the subflow that creates the credit without a reason, and when it fails.

06Org evidence

Facts only the org has, from your own CLI.

Metadata shows what exists. It doesn't show who holds a permission set, whether a rule is already failing, or how full storage is. For those, the AI asks, and your CLI agent answers.

01The AI writes the questions: When the metadata can't answer something, such as who holds a permission set or whether a rule is already failing, the analysis pauses with a brief: each question, why it matters, and the checks to run.
02Your CLI agent runs them: Copy the brief into Claude Code or Codex in your Salesforce project. The agent runs read-only queries against your org and writes a findings packet into the project folder.
03The packet appears by itself: In Chrome or Edge, with the workspace folder connected, the packet shows up on the request a few seconds after the agent writes it. Elsewhere, import it from the folder. Each packet says which org its queries ran against.

symposiumai.app/analyze

An org-limit run asks for storage readings. The packet the CLI agent writes appears on the request by itself, and the forecast follows.

The kit

One download and one command in your Salesforce project. It installs a Claude Code skill that answers Symposium's requests, rules for Codex, a guard that blocks deploys and data changes, and a validator that checks a packet the way Symposium does.

The setup steps for the CLI kit: download symposium-cli-kit.zip, unzip it, and run the installer in the Salesforce project folder — Setting up the kit: download it, unzip it, run the installer in your project.

07The whole workspace

All of your metadata, with packages kept apart.

Every metadata type in your package directories, in source or metadata format. Our demo org's 34,610 metadata files upload as a 21 MB bundle.

Installed packages are inventory, not risk. The audit lists each package with what it contains and counts your own metadata apart, so its findings are about what you can change.
Python runs over every file, for the questions a built-in audit doesn’t answer.
Each analysis says which org and which pull it describes.

A workspace audit report: a summary of logic conflicts between automations, and headline figures for bookings that would fail to cancel, re-rated signals and price-lock coverage — An audit report on the demo org: what to fix first, with each figure labeled by where it came from.

08Who it's for

Anyone who has to understand an org they didn't build.

Admins: Understand an org you inherited, and see who can see what before an access review.
Architects: Draw the data model and the automation on an object, and check what a change would reach.
Developers: Trace an error to the Flow element or line of Apex behind it, with the documentation that explains why.
Release managers: See what a seasonal release or a planned feature touches, with a checklist.
Consultants and partners: Get to know a client's org fast, and hand off work as acceptance criteria and prompts.
Education orgs: An access inventory for sensitive fields, such as student records: which profiles and permission sets can read or edit each one.

09Your data

What's uploaded, and where it lives.

The workspace bundle and anything the AI produces run in a private, isolated sandbox for your session, hosted by our code-execution provider E2B, with no internet access.
Uploads are also kept with the session in our cloud storage, so they are restored if the sandbox is recycled. Sandboxes are deleted after 7 days of inactivity, when you delete the session, or when you delete your account; deleting the session also deletes its uploads.
Your AI provider keys are stored AES-256 encrypted on our servers and never sent to your browser. Symposium never asks for, or stores, Salesforce credentials.

The details are in the privacy policy.

10FAQ

Salesforce questions.

Does it connect to my org?

No. You upload metadata you already pulled, as a folder or a zip. When an analysis needs facts only the org has, your own Claude Code or Codex gathers them read-only, on your machine. Symposium never holds org credentials.

Which formats does it read?

Source format (a Salesforce DX project with sfdx-project.json) or metadata format (a retrieve with package.xml). It takes every metadata type in your package directories.

Do I need Claude Code?

No. Everything except org evidence works from the uploaded metadata alone. For org evidence, Claude Code and Codex both work, and you can always continue without it.

Can it change my org?

No. Symposium never deploys, writes data, or edits your source. Handoffs are acceptance criteria and implementation prompts, never patches. The CLI kit adds a guard that blocks deploys and data changes, and if your project's CLAUDE.md or AGENTS.md tells agents to deploy, Symposium warns you before you hand anything off.

Which AI providers can run it?

Any provider with function calling: Claude, ChatGPT, Gemini, Grok, Mistral, Cohere, Kimi, or GLM, with your own API keys.

More about Analyze in the FAQs.

Point it at a workspace you already have.

Analyze a workspace

Salesforce, Apex and Flow Builder are trademarks of Salesforce, Inc. Claude Code is made by Anthropic and Codex by OpenAI. Symposium AI is not affiliated with or endorsed by any of them. Screenshots show a demo org built from Salesforce's Coral Cloud sample app.