One AI agent runs yourSalesforce integrations.
It works natively inside Salesforce and reaches every tool around it over the Model Context Protocol. The same agent reads the requirement from an email, the spec from a doc, the bug from a ticket, then builds and deploys the fix in your org.
Builds + deploys in
Your Salesforce org
How it works
Two layers, one agent.
The Clientell Salesforce integration works on two layers:
Inside your org: the agent operates natively across metadata, flows, Apex, reports, permissions, data, and deployments, with no middleware and no IDE to learn.
Around your org: it connects to the tools your work lives in, Gmail, Outlook, Google Drive, Dropbox, SharePoint, GitHub, Jira, and Microsoft Teams, over the Model Context Protocol (MCP), the open standard Anthropic introduced in 2024 and Salesforce adopted in 2025.
On your terms: connect what you want with your own scoped login, or connect nothing and it still runs on Salesforce out of the box.
An open standard
Nov 2024 → Jun 2025MCP, from Anthropic, now adopted across Salesforce. Built for admins, no IDE.
Works out of the box
Runs on your org with nothing connected. Attach tools per conversation with your own scoped login.
Audits what you already run
Inventories connected packages and email services, with known vendors labeled.
Read-only by default
Scoped login, tokens isolated per user and redacted from logs.
01How it connects
Native inside Salesforce. MCP for everything around it.
Inside your org
Native Salesforce
No middleware. The agent acts directly through the Salesforce REST, SOQL, Tooling, Metadata, and Bulk APIs.
Around your org
MCP connectors
Over the Model Context Protocol, with your own scoped login. Read-only by default, attached per conversation.
02Inside the org
The deepest integration is Salesforce itself.
Metadata & config
Objects, fields, page layouts, list views, validation and duplicate rules, created and updated through the Metadata and Tooling APIs.
Flows
Build flows from plain English, validate them against your live org, and deploy. Broken subflows and dangling references get repaired before deploy.
Apex & code
Triggers, classes, LWC, and Visualforce, plus anonymous Apex and generated test classes, written governor-safe.
Reports & dashboards
Create, clone, and update reports (summary, matrix, joined) and build dashboards, with scheduled refresh.
Permissions & users
Field-level security, permission sets, profile cloning, and password resets, traced to the source.
Data operations
SOQL, CRUD and upsert, the Bulk API for large jobs, plus dedupe, merge, and field-fill.
Deployments
Sandbox to production with a check-only validation first, a full diff, and one-click rollback from a pre-deploy snapshot.
Org audit
Your org graded across configuration, automation, tech debt, compliance, packages, adoption, and AI readiness.
Dependency graph
Ask where a field is used or what a change would break, answered from a live graph of objects, flows, Apex, and reports.
Scheduled jobs & change log
Schedule Apex, list and abort jobs, and review a 180-day record of who changed what.
Works in sandbox and production. Validated before deploy, with one-click rollback.
03The connectors
Ten tools, connected over MCP.
Gmail
Read-onlySearch and read email threads for the context behind a task.
Outlook Email
Read-onlySearch and read Outlook mail as a source for the work.
Google Calendar
Read-onlyPull meeting and scheduling context into the task.
Outlook Calendar
Read-onlyReference Outlook events alongside the org.
Google Drive
Read-onlyPoint the agent at a doc or sheet as the spec for a build.
Dropbox
Read-onlyUse files in Dropbox as input for the task.
SharePoint
Read-onlyRead documents across SharePoint sites and libraries.
GitHub
Read-onlyRead repos, code, issues, and PRs. Writes for Salesforce code go through a controlled deploy path.
Jira
Read + writeRead issues and projects, and create issues, add comments, or transition status.
Microsoft Teams
Read-onlySearch messages and read threads where your team works.
Plus a Slack-native interface: the agent runs in a Slack thread with the same Salesforce and connector reach it has on the web. Connector access reflects the shipped providers as of June 2026 and can change.
04What is Salesforce MCP
The open standard, built for admins
An open standard: MCP (the Model Context Protocol) connects AI agents to external tools and data. Anthropic introduced it in November 2024, OpenAI, Google, and Microsoft adopted it through 2025, Salesforce shipped MCP support across the platform in June 2025, and Anthropic donated MCP to the Linux Foundation in December 2025. There are now more than 10,000 public MCP servers.
Clientell is the MCP host: the agent reaches each tool through that tool's own MCP server or API, using your scoped OAuth token. You attach the connectors you want per conversation, tokens are isolated per workspace and user and redacted from logs, and the file, email, and calendar connectors stay read-only.
Built for admins, not developers: you can connect a generic AI like Claude to Salesforce over MCP, but it assumes a developer in an IDE, managing API billing and prompts, with no Salesforce guardrails. Clientell gives admins the same reach without any of that. ChatGPT tells you what to do. Clientell does it, inside your org.
MCP, in order
Nov 2024
Anthropic introduces the Model Context Protocol as an open standard.
2025
OpenAI, Google, and Microsoft adopt MCP.
Jun 2025
Salesforce ships MCP support across the platform.
Dec 2025
Anthropic donates MCP to the Linux Foundation. 10,000+ public servers.
05Why admins choose Clientell
Generic AI over MCP vs Clientell
For comparison on cost: Agentforce lists at $125 per user per month and needs Data Cloud, before implementation. Clientell starts at $99 per month, flat, and needs neither.
06The integrations you already have
Find the one that broke quietly
The problem: most orgs don't have an integration problem because they have too few. They have one because nobody can see all the ones they already run.
What it inventories: the connected apps, your org-wide email addresses and email services, and every installed package.
What it surfaces: it labels known vendors as it goes, Outreach, Gong, SalesLoft, ZoomInfo, DocuSign, and the rest, so you get a real map instead of a list of cryptic package names: the sync that stopped writing, the duplicate email service, or the package nobody remembers installing.
It is also one of the areas graded in a full org audit.
Illustrative readout
07In practice
Salesforce work rarely starts in Salesforce.
The requirement is in an email
Point the agent at the thread, and it builds what the stakeholder actually asked for instead of what you retyped.
The spec is in a doc
Reference a Drive, Dropbox, or SharePoint file as the source, and the agent builds the flow or report to match.
The bug is in a ticket
Bring a Jira or GitHub item into the work, fix it in the org, and update the ticket when it's done.
The ask comes in Slack
Run the whole thing inside Slack. The agent has the same Salesforce and connector reach in a thread as it does on the web.
Custom & managed integrations
Need Salesforce talking to NetSuite or HubSpot?
System-to-system integrations are part of our AI-led services. We scope the data model, build the sync, and run it, at a fixed price instead of an open-ended hourly engagement.
Questions, answered.
What does Clientell integrate with?
Two layers. Inside your org, the agent works natively across metadata, flows, Apex, reports, permissions, data, and deployments. Around your org, it connects to ten tools over the Model Context Protocol: Gmail, Outlook email, Google and Outlook Calendar, Google Drive, Dropbox, SharePoint, GitHub, Jira, and Microsoft Teams. It also runs natively inside Slack.
What is Salesforce MCP?
MCP, the Model Context Protocol, is an open standard for connecting AI agents to external tools and data. Anthropic introduced it in November 2024, OpenAI, Google, and Microsoft adopted it through 2025, and Salesforce shipped MCP support across the platform in June 2025. Clientell acts as an MCP host, so the agent reaches your connected tools over the same standard Salesforce itself now runs on.
Do I need an IDE, Data Cloud, or Agentforce to use this?
No. You can technically connect a generic AI to Salesforce over MCP, but it assumes a developer in an IDE managing API tokens and prompts. Clientell gives admins the same cross-system reach with no IDE, no Data Cloud, and flat pricing from $99 per month.
Is connecting a tool secure?
You connect each tool with your own scoped OAuth login, and you control what is attached, per conversation. The file, email, and calendar connectors are read-only. Tokens are isolated per workspace and user and redacted from logs. Clientell is SOC 2 Type II, HIPAA, and GDPR compliant.
Can the agent write back to my connected tools?
The email, calendar, and file connectors are read-only by design: they bring context in, they don't change your inbox or drive. Jira is the exception, where the agent can create issues, add comments, and transition status. Writing code to GitHub happens only for Salesforce work, through a controlled deploy path.
Can it audit the integrations I already have?
Yes. The agent inventories the connected apps already in your org, your org-wide email addresses, email services, and installed packages, and labels known vendors like Outreach, Gong, SalesLoft, ZoomInfo, and DocuSign. That is how you find the integration that broke quietly.
Do I have to connect anything to get started?
No. The agent works on your Salesforce org out of the box. Connectors are optional, added when a task needs context from outside the org.
Can you build a custom integration, like NetSuite or HubSpot?
Yes. System-to-system integrations are part of our AI-led services. If you need a custom or managed integration between Salesforce and another platform, we scope and ship it at a fixed price. Book a consult and we'll map it.
Sources
- Anthropic: Introducing the Model Context Protocol · November 25, 2024
- Salesforce Developers: Introducing MCP Support Across Salesforce · June 23, 2025
- Anthropic: Donating MCP to the Linux Foundation · December 9, 2025
MCP adoption and connector availability reflect public information as of June 2026 and can change. Verify current provider support in the Clientell app.
Keep exploring
Connect your stack. Let one agent work across all of it.
The agent works on Salesforce out of the box. Connect email, docs, tickets, and chat over MCP when a task needs the full picture.