# Ask Claude about your Salesforce org, and get answers from the org itself.

> The SyncOnAI 360 Claude connector lets Claude read your connected Salesforce orgs through a Model Context Protocol server. Claude can list the orgs a token allows, search their metadata, look up a component and everything that depends on it, and read health findings. Access is read-only, scoped per token, logged and revocable immediately.

Source: https://synconai360.com/connectors/claude

## Key facts

- **2 min**: To connect Claude Desktop
- **0**: Write actions available to Claude
- **120**: Calls a minute allowed per token
- **Every**: Call logged with its tool and token

## Why Claude on its own gets Salesforce questions wrong

Claude knows Salesforce in general. It does not know your org: which fields exist, which Flows run on Case, what a validation rule checks. Without that, it answers confidently about an org it has never seen.

- **Generic answers.** Claude describes how Salesforce usually works, not how your org is built.
- **Copy and paste.** People paste metadata into the conversation, which is slow, partial and leaves no record of what was shared.
- **Too much access.** Giving an AI client a Salesforce login hands it the power to change the org, which few teams want.
- **No accountability.** Nobody can say afterwards which AI tool read what, or when.

## What can Claude do with your Salesforce org?

Through the connector, Claude can list the orgs its token may see, search their metadata, look up a component and everything that depends on it, and read the findings of a health assessment. Ask in production org, what uses the Account object, and Claude answers from the synced repository rather than from general knowledge.

Every answer says how much of the org it was based on. If a search hit its limit, or the token can see only some orgs, Claude is told, so a negative answer like nothing uses this field can be trusted or questioned for the right reason.

- List orgs and search metadata
- Look up dependencies
- Read assessment findings
- Coverage stated with answers

## Why is the Claude connector read-only?

Changes to an org go through a proposal in SyncOnAI 360, where they are checked against the org and approved before they deploy. A token held by an AI client has none of those checks behind it, so it is never given the ability to write.

Claude can create, edit, deploy and delete nothing in a Salesforce org. When an answer leads to a change, make it in SyncOnAI 360, where it becomes a proposal with named checks, validation and approval.

- No write actions at all
- Changes go through proposals
- Approval stays with your admins

## How do you connect Claude Desktop to Salesforce?

In Settings, under Integrations, create an access token, name it after where it lives, and choose which orgs it may read. Copy it straight away; only a one-way digest of it is stored, so it cannot be shown again.

In Claude Desktop, open Settings, then Developer, then Edit Config, and add the SyncOnAI 360 server address with the token as a bearer header. Restart Claude Desktop and the connector appears in the tools menu. Any other Claude client that speaks Model Context Protocol over HTTP connects the same way.

- Create a scoped token
- Add one block to the config
- Restart and ask

## How do you control and audit Claude's access?

Each token is scoped to the orgs you choose and can be revoked in Settings, where it stops working immediately. Use one token per client, so revoking a lost laptop's token does not disturb anyone else.

Every call Claude makes is written to the activity log with the tool it used and the token it used, and the settings page shows when each token was last used. The arguments are not recorded, so the log is accountability for access rather than a second copy of your org.

- Tokens scoped per org
- Instant revocation
- Every call logged
- Last used shown per token

## How does this differ from the AI inside SyncOnAI 360?

The chat inside SyncOnAI 360 reads the same org and can also draft changes, which become proposals with checks and approval. The Claude connector is for people who prefer to work in Claude and want their org available there, read-only.

Both read the same synced copy of the org, so an answer in Claude and an answer in the workspace agree with each other.

- Same synced org behind both
- Drafting changes stays in the workspace
- Read-only in Claude

## Connecting Claude

1. **Create a token.** Name it and choose the orgs it may read.
2. **Add the server.** Paste the address and token into Claude's config.
3. **Restart.** The connector appears in Claude's tools menu.
4. **Ask.** Which orgs can you see? Then something real.

## Claude and Salesforce, pasted versus connected

| Without | With SyncOnAI 360 |
|---|---|
| Metadata pasted into a conversation | Claude reads the synced org directly |
| General answers about Salesforce | Answers about your org, with coverage stated |
| A login that could change the org | Read-only access by design |
| No record of what was read | Every call logged with tool and token |
| Access hard to take back | Revoke a token and it stops at once |

## Who connects Claude

- **Salesforce architects.** Explore an org's structure and dependencies in a long Claude conversation, with every answer drawn from the org itself.
- **Salesforce admins.** Ask what uses a field or what an org contains without leaving the Claude window you already work in.
- **Consultants.** Give Claude read access to a client's org, scoped to that org only, and revoke it when the engagement ends.
- **IT and security leads.** Let people use Claude on Salesforce with read-only, scoped and logged access instead of pasted extracts.

## Frequently asked questions

### Can Claude change my Salesforce org?

No. The connector is read-only. Changes go through a proposal in SyncOnAI 360, where they are checked and approved.

### Which Claude apps work?

Claude Desktop, and any Claude client that speaks Model Context Protocol over HTTP.

### What can Claude read?

The orgs its token allows: their metadata, the components that depend on each other, and health findings. Not CRM records.

### Does Claude see our customer records?

No. SyncOnAI 360 keeps org metadata, not CRM records, and that is all the connector can read.

### How do we limit which orgs Claude sees?

Choose the orgs when creating the token. Leave every box unticked to allow all of them, or tick specific orgs.

### What if a token is lost?

Revoke it in Settings and it stops working immediately. Creating a replacement takes seconds.

### Is Claude's access logged?

Yes. Every call is recorded with the tool and the token used. The arguments are not recorded.

### Is there a rate limit?

120 calls a minute per token, well above what one question needs.

### Is the Claude connector on every plan?

Yes. Every plan includes every feature.