Every Monday, we pull back the curtain on what’s shipping in Workato Enterprise MCP.
This Week: Three New Servers and a GitHub Upgrade
Update — March 2026: People.ai is now in production with the Salesforce Sales Cloud Explorer MCP Server featured in this post. They used it to give AI agents governed, scoped access to live production CRM data — and wrapped Apollo.io through Workato Enterprise MCP to add enrichment in the same governed workflow. The result: 350+ existing automation recipes became Enterprise Skills agents could invoke directly, and AI-driven email workflows saw a 20% lift in open rates. Read the full story →
Three new pre-built MCP servers are live today — Gmail, Salesforce Sales Cloud Explorer, and Airtable. We also shipped a significant upgrade to the GitHub server, adding full write capabilities.
This week’s theme: agents that can take action where revenue, code, and customer operations actually happen.
What Shipped
Gmail — 18 tools Complete email access with a safety-first design. Search threads, retrieve full context, draft emails, and send — but only after explicit user approval. No direct send, no permanent deletion.
View Gmail documentation →
Salesforce Sales Cloud Explorer — 15 tools Full CRUD across Sales Cloud core objects — leads, accounts, contacts, opportunities, tasks, events, and products. Plus natural-language SOQL for custom field queries. Users describe what they want in English. The agent writes SOQL.
View Salesforce Explorer documentation →
Airtable — 11 tools Deterministic, schema-aware access to Airtable bases used as operational systems of record. Read from curated views, create and update records, batch operations, and full schema discovery.
View Airtable documentation →
GitHub — Now with Write Capabilities — 13 tools (5 new) The GitHub server now supports full CRUD on issues and pull requests. Agents can create issues, update assignees and labels, add comments, and open PRs — not just read. No PR merging or approval (that still requires human judgment).
View GitHub documentation →
Why This Matters
These releases cross a critical threshold: agents can now read and write to the systems where revenue, code, and customer relationships are managed.
An agent can search your Salesforce pipeline for stalled deals, pull the latest email thread with that customer from Gmail, update account health in Airtable, and create a GitHub issue for the engineering blocker — all in one conversation. That’s not a demo. That’s how work actually flows.
Real-World Use Cases
Gmail: For Customer-Facing Teams
- Morning triage: “Search my inbox for customer emails from the last 24 hours. Flag anything mentioning escalation, cancel, or disappointed.”
- Contextual responses: “Pull up the full thread with DataVault’s VP. Draft a reply addressing their API concerns.”
- Follow-up tracking: “Find all emails I sent last week that haven’t gotten a response.”
- Inbox management: “Archive everything from the conference threads. Label the partner emails as follow-up.”
The safety model: Every email starts as a draft. The agent creates it, you review it, you approve the send. No autonomous email. No permanent deletion — archive over delete. This is the most cautious write model in our server library, by design.
Salesforce Explorer: For Revenue Teams
- Pipeline inspection: “Show me all opportunities closing this quarter over $50K that haven’t changed stage in 30 days.”
- Real-time updates: “Update the Meridian deal to Negotiation, $175K, close end of next month.”
- Lead capture: “Create leads for the three contacts I met at the AWS conference. Check if their companies already exist in Salesforce.”
- Account intelligence: “Pull the account hierarchy for GlobalTech. Who are the contacts? Any open tasks?”
- Custom queries: “Show me all opportunities where our custom ‘competitor_mentioned’ field equals Snowflake.
Customer Spotlight: People.ai deployed this server to give AI agents access to live Salesforce data without waiting for vendor MCP support — and used the scoped query architecture to avoid pushing against LLM token limits on large datasets. See how they did it →
The SOQL unlock: The Explorer includes retrieve_semantic_model and execute_soql_query — agents can discover your custom fields and objects, then query them through natural language. Nobody needs to learn query syntax.
Airtable: For CS and Operations Teams
- Account reviews: “Pull all accounts flagged high-risk from the Q1 Renewals view.”
- Operational updates: “Update TechVault’s risk level to medium. Mark DataStream’s milestone as complete. Add a new blocker for Meridian: procurement delay.”
- Board prep: “Show me customer health data from the accounts base. Cross-reference with the milestones tracker.”
- Cleanup: “Delete all resolved blockers from Q4.”
The view-first pattern: If your team has built curated Airtable views with specific filters and sorting, agents read from those views directly — respecting the business logic your team already defined. The agent doesn’t reconstruct your filters with raw queries. It trusts your views.
GitHub: Now with Write Capabilities
- Issue creation from conversation: “Create a bug for the API rate limit issue we just discussed. Assign to Sarah, label it P1.”
- PR management: “Open a PR from the feature/auth-v2 branch. Add Marcus and Priya as reviewers.”
- Triage at speed: “Update all three open bugs from today’s standup — assign the auth issue to Dev, the cache issue to Sarah, and close the logging issue as won’t-fix.”
- Comment without context-switching: “Add a comment on issue #347: ‘Root cause identified — fix incoming in PR #351.'”
What’s new: Five new tools — create_issue, update_issue, add_issue_comment, create_pull_request, and update_pull_request. The original 8 read tools remain unchanged. Same repo search, same commit history, same PR review — now with the ability to act on what you find.
What Makes These Enterprise-Ready
- Verified User Access: Gmail, Salesforce, and Airtable execute under the authenticated end-user’s identity. GitHub uses authenticated tokens. Every action is traceable to a specific person.
- Draft-first sending (Gmail): Agents can’t send email directly. Every email is created as a draft and requires explicit user approval.
- Read-first, confirm-before-write: All servers retrieve current state before modifying anything.
- Managed infrastructure: Workato hosts, scales, and updates. No servers to manage, no API versions to track.
- Composable: Chain tools across any combination of pre-built servers for complete workflows.
- Audit trails: Every action logged with full context — who, what, when, why.
Get Started
All three new servers and the GitHub upgrade are available now for Workato customers with Enterprise MCP.
Install time: Minutes
Custom development: None
Deployment complexity: One-click activation
View Gmail documentation →
View Salesforce Explorer documentation →
View Airtable documentation →
View GitHub documentation →
The bottom line: Other vendors are announcing MCP roadmaps. We’re shipping production-ready servers every Monday.
See you next Monday.