# Bring your org into the AI tools your team already uses.

> SyncOnAI 360 includes a Model Context Protocol server that lets Claude Desktop, Cursor and other MCP clients read your connected Salesforce orgs: list them, search metadata, see what depends on a component and read assessment findings. Access is read-only, scoped per token, logged, and revocable instantly.

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

## Key facts

- **2 min**: To connect Claude Desktop or Cursor
- **0**: Write actions available to an external AI client
- **120**: Calls a minute per token, well above normal use
- **Every**: Call logged with the tool and the token

## Why connecting AI tools to Salesforce worries security teams

Developers want their AI assistant to understand the org. Security teams see a token that can change production sitting in a config file on a laptop. Most integrations force a choice between useful and safe.

- **Tokens with too much power.** A Salesforce connection that can write gives any tool holding it the ability to change production, with no review and no approval.
- **No scoping.** A token for one developer's experiment can see every org the workspace has, including the regulated ones.
- **No record.** When an AI tool reads an org, nobody knows what it asked for, when, or with which credentials.
- **Negative answers you cannot trust.** An assistant that says nothing uses a field may simply have hit a search limit. Without coverage, a negative answer is a guess.

## What can an MCP client do with SyncOnAI 360?

An AI client can list the orgs its token allows, search their metadata, look up a component and everything that depends on it, and read the findings from a health assessment. Ask Claude "In the production org, what uses the Account object?" and it answers from your org.

Every answer says how much of the org it was based on, which matters most for negative answers: the client is told when a search hit its limit or when its token could only see some orgs. An org that has never been assessed is reported as exactly that, not as an empty list of findings.

- List orgs, search metadata, find dependencies, read findings
- Coverage stated with every answer
- Works with Claude Desktop, Cursor and any MCP client over HTTP

## Is the Salesforce MCP connection read-only?

Yes. An external AI client cannot create, edit, deploy or delete anything in a Salesforce org. Changes go through proposals inside SyncOnAI 360, where they are validated and approved; a client holding a token has none of that, so it is not given the ability to write.

Each token is scoped to the orgs you choose, and it is stored in a form that cannot be read back, so copy it when it is created. Revoke a token and it stops working immediately. Tokens are limited to 120 calls a minute, well above normal use and well below anything that could hammer the workspace.

- No create, edit, deploy or delete
- Tokens scoped to chosen orgs
- Instant revocation, one token per client
- Rate limited per token

## How is AI client access recorded?

Every call an AI client makes is written to your workspace's 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. A search term can describe the shape of your org, and the purpose of the record is accountability for access, not a second copy of what was accessed.

- Tool and token recorded for every call
- Last-used time for every token
- Arguments deliberately not stored

## Which delivery tools does SyncOnAI 360 integrate with?

Jira: raise audit findings as Jira Cloud issues, with status checked back on demand. GitHub: record each deployed change as a pull request in a repository you choose. Slack: post every deploy result to a channel. Jira and GitHub are in beta.

Project webhooks send work item events, created, updated, status changed, comment added or attachment added, to any other system you run. Admins set each integration up once for the workspace.

- Findings raised as Jira issues
- GitHub repositories connected to the workspace
- Slack notifications for deploys and audits
- Project webhooks for everything else

## Can the SyncOnAI 360 chat use other companies' tools?

Yes. Alongside the tokens your own AI clients use to read your orgs, a workspace can add another company's tool server, such as Atlassian for Jira and Confluence, and the chat can use its tools. Use Test to check a saved connection before relying on it.

Every call the chat makes to an outside tool is recorded in the activity log, the same way calls from external AI clients are. Connections are managed by admins in the workspace's MCP settings, so what the chat can reach is a decision, not a default.

- Add tool servers such as Atlassian for Jira and Confluence
- Test a connection before using it
- Every outside tool call recorded

## How to connect an AI client

1. **Create a token.** In Settings, name the token after where it will live and choose which orgs it can read.
2. **Copy it.** Copy the token straight away; it cannot be shown again.
3. **Add it to your client.** Add the SyncOnAI 360 address and the token to Claude Desktop, Cursor or another MCP client.
4. **Ask.** Ask which orgs it can see, then ask a real question about one.

## AI tools on Salesforce, with and without SyncOnAI 360

| Without | With SyncOnAI 360 |
|---|---|
| Tokens that can change production | Read-only access with no write actions |
| Every org visible to every tool | Tokens scoped to the orgs you choose |
| No record of AI access | Every call logged with tool and token |
| Negative answers you cannot trust | Coverage stated with every answer |
| Delivery tools disconnected from the org | Findings in Jira, deploys in GitHub and Slack |

## Who connects through MCP

- **Salesforce developers.** Give Cursor or Claude Code the org's real model, so generated code uses real fields and knows its dependents.
- **Salesforce architects.** Explore dependencies across an org from Claude, with every answer drawn from the synced org and its coverage stated.
- **Consultancies.** Scope a token to one client's org for each consultant, and revoke it the day the engagement ends.
- **IT and security leads.** Allow AI clients on Salesforce with read-only, scoped and logged access instead of pasted extracts.

## Frequently asked questions

### What is a Salesforce MCP server?

A Model Context Protocol server lets AI clients such as Claude Desktop and Cursor call tools to read data. SyncOnAI 360's MCP server lets them read your connected Salesforce orgs' metadata, dependencies and assessment findings.

### Can Claude or Cursor change our Salesforce org?

No. The connection is read-only. An external AI client cannot create, edit, deploy or delete anything. Changes go through validated, approved proposals inside SyncOnAI 360.

### Which AI clients are supported?

Claude Desktop, Cursor and any other client that speaks Model Context Protocol over HTTP, using a bearer token.

### Can we limit which orgs a token can see?

Yes. Choose specific orgs when creating a token, or leave it open to all. Prefer narrowly scoped tokens.

### What happens if a token is lost?

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

### Is AI client access logged?

Yes. Every call is written to the activity log with the tool and the token used. Arguments are not recorded.

### Are there usage limits?

Each token is allowed 120 calls a minute. One question typically takes three or four calls.

### Does SyncOnAI 360 integrate with Jira?

Yes, in beta. Audit findings can be raised as Jira Cloud issues with their evidence, and status is checked back on demand. Admins connect Jira once for the workspace.

### Can we get deploy notifications in Slack?

Yes. After connecting Slack, configure channels for deploy outcomes and for audit completion.

### Does the connection work with Claude Code and other agents?

Any client that speaks Model Context Protocol over HTTP and can send a bearer token can connect, using the address and token from your workspace settings.

### Is the Jira integration two-way sync?

No. It raises findings as Jira issues and checks their status on demand. It is not a two-way synchronisation of tickets.

### Is there a Slack bot?

Not as a conversational bot. Slack receives notifications, such as deploy outcomes and audit completion, in the channels you choose.

### Is the MCP server included on the free plan?

Every plan includes every feature. Free covers one user and one production org with its sandboxes, on your own Anthropic or OpenAI key, with no credit card and no time limit. Paid plans include AI and cover more people and production orgs.

### What should we do with tokens when someone leaves?

Revoke their tokens in Settings. A token is a password: anyone holding it can read the orgs it is scoped to. Create one token per client so revoking one does not disturb the others.