Integrations/SalesforceLast updated
Integration

Salesforce AI integration for pipeline answers and record updates

Coworker's Salesforce AI integration lets anyone ask about accounts, opportunities and cases in plain English and get answers from live Salesforce data, next to your calls, email and Slack. Agents can run a saved report, log the task or update the opportunity. Changes can wait for your approval, so you see every field before it updates.

How Salesforce context reaches your AI tools

Coworker brings Salesforce together with the rest of your tools in OM2, its organizational memory, then answers in Coworker's own apps, or in Claude, ChatGPT and Cursor through Coworker MCP.

Use cases

What teams do with the Salesforce AI integration

Coworker's Salesforce AI agents read and update Salesforce next to your calls, email and Slack. Here is what each team uses them for.

RevOps

Pipeline numbers on request

Ask for open pipeline by stage and owner and get a report from live Salesforce data, across standard and custom objects.

RevOps

The reports you already trust

Point Coworker at a saved Salesforce report and get a summary built from the same rows your team already reviews.

Sales leaders

Deal reviews

See which opportunities in Negotiation have gone quiet for 14 days, with each record next to its last call and email thread.

Account executives

Tasks and next steps

After a call, log a task on the opportunity and update the next step and close date. Each write comes back with a View in Salesforce link.

Admins

Updates that fit your org

Coworker checks every update against your org's fields before it runs, so changes match how your Salesforce is set up. A value outside a restricted picklist is held back with the allowed options shown, and missing required fields are caught first.

More for revenue teams: Coworker for sales teams, and our Salesforce MCP guide.

Setup

Live in minutes

1

An admin connects Salesforce

An admin connects your Salesforce org once for the workspace and chooses who can search it.

2

Each person adds their own login

Writes and report runs use each person's own Salesforce login, so people can only change what Salesforce already lets them change.

3

Decide what needs approval

Tick Require approval for Salesforce on an agent. Each change then shows up as an editable card and only runs when someone approves it.

Powered by the Coworker Platform / OM2

Every model needs the right context.

Salesforce's MCP servers can only see what lives in Salesforce. OM2 connects that data to the Slack threads, Jira tickets, Gong calls, and Google Drive docs around it in one permission-aware knowledge graph, so Claude, ChatGPT, or Cursor starts every task with your customers, projects, and decisions already loaded.

customer14s ago

Mosaic Financial

Series B fintech · $3.1M ARR · 47 seats

24linked
12sources
0.8ssynth
SOC 2 Type II auditPriya NambiarEnterprise onboardingCustomer Success#mosaic-impl
LiveOnboarding kickoff call with Priyavia Gong
Security and governance

Enterprise-grade security, from day one

SOC 2 Type II

Audited annually, report on request

Hosted locally

Your memory stays in your region. Open-source models run in the US.

Isolated per company

In our cloud or your VPC

Never trains on your data

No model provider, closed or open source

Your permissions, enforced

The access already set in Salesforce, Slack, Jira and Drive

Also included

GDPRCASA Tier 2Node-level permissions at query timeEvery fact traceable to its sourceSSO and SAMLAES-256 at rest, TLS 1.2+ in transitRegional data residency

FAQ

Frequently asked questions

Coworker is a context layer for enterprise AI. It connects 50+ work tools, builds a permission-aware company graph called OM2, and runs agents that act across those tools, inside Coworker or in Claude, ChatGPT and Cursor through Coworker MCP.

Coworker connects Claude, ChatGPT and Cursor to Salesforce through Coworker MCP, one connection that also brings in Slack, Jira, Google Drive, call transcripts and the rest of Coworker's 50+ connectors. Salesforce also offers its own hosted MCP servers for Enterprise Edition orgs, which cover Salesforce data. Our Salesforce MCP guide compares the two.

Salesforce's hosted MCP servers are a strong fit if you only need Salesforce data in Claude: they are generally available on Enterprise Edition and above and apply your sharing rules to every call. Coworker MCP is for answers that also need Slack, Jira, Gong calls or Google Drive, in one connection with Salesforce and the rest of Coworker's 50+ connectors.

Agentforce is Salesforce's own agent platform, grounded in Salesforce records and Data Cloud, with MuleSoft to reach other systems. If your team builds on the Salesforce platform, it is a strong fit. Coworker reads Salesforce next to Slack, Jira, call transcripts and docs through its own 50+ connectors, and Salesforce is one of the sources its organizational memory draws on.

Any standard or custom object, such as accounts, contacts, opportunities, cases and activities. Ask in plain English and Coworker writes the query for you. It can also look up an object's fields, picklist values and record types, and run your saved reports.

With a person's own Salesforce login connected, agents can create and update records on standard and custom objects, including tasks, events and notes. They can also delete a record when asked. Without that login, Coworker only reads.

You decide. Tick Require approval for Salesforce on an agent and every write from it waits for a person. The approval card shows the record and each field, which you can edit before anything runs.

Coworker checks every update against your org's fields before it runs, so one bad value never sinks the whole change. Values outside a restricted picklist are held back with the allowed options shown, a required field with a bad value stops the update before it reaches Salesforce, and the approval card shows anything that will be left out.

Admins decide who can search Salesforce through Coworker, and anyone can connect their own Salesforce login. Every change and report run uses the person's own login, so Salesforce's permissions decide what they can edit.

Yes. An agent can run on a schedule, when a meeting ends, or when someone asks in Slack, and send its results to Slack or email. For example, a Monday pipeline summary, or a weekly list of opportunities past their close date.

Coworker is an enterprise AI assistant whose agents act in all three. In Salesforce, Coworker creates and updates records and logs tasks. In Jira, it files issues, comments and moves them through the workflow. In Google Drive, it creates Docs, Sheets and decks. Every update can wait for a person's approval before it runs.

Connect Salesforce in minutes

Connect your Salesforce data and watch agents go to work in minutes.