EP03: Salesforce Developer AI Agent is OUT!
Built an AI assistant for devs that writes:
• Apex triggers
• LWCs
• Test classes (95%+)
Not replacing devs - AI handles repetitive coding while you control.
🎥 https://t.co/ym027kVxXh
@salesforce#claudecode#salesforcedxmcp
One command gives Claude Code about 40 Salesforce development skills, three MCP servers, an agent, hooks and commands. Here's the setup and a real first prompt in my dev org: https://t.co/I9jy0KSohn
Salesforce says personal AI agents will drive more contact center traffic than they offset. They wait on hold, call back every few hours and keep no business hours. How to meet them: https://t.co/n2jvGPmqYY
Salesforce signed a definitive agreement to acquire Cresta, whose voice agents already take complex calls for Marriott, Comcast and Alaska Airlines. It isn't closed yet. What it means for Service Cloud: https://t.co/UpsSbiTrdF
Utah gave Claude Code to 34 developers across 10 state agencies. 71% reported saving 20+ hours in a four-week pilot (self-reported). How I'd run the same pilot with a Salesforce team: https://t.co/cjKXEdTuO1
Agent Optimizer can pick your best use case from real cases, reuse your flows and actions, and test the agent until it passes a quality bar. It's in beta. My take: clean case data is now agent prep. https://t.co/nHmDw96Kbo
Your Salesforce flows can finally have real regression tests. Flow Test Mode (beta) adds saved scenarios, assertions and mock outputs, and Agentforce can write the tests from your terminal. https://t.co/I1KsUsSHmC
A Claude Code subagent limited to Read, Grep and Glob can review your Apex without changing a line. It checks for SOQL in loops, missing sharing keywords and hardcoded IDs. The setup: https://t.co/Kppyx26RsL
Salesforce added a kill switch for AI agents. Agent Fabric now has hard spending limits that stop requests before costs pass budget, plus an Agent Kill Switch. Agent Evals goes into beta this month. https://t.co/pV1SgY4Olo
A Claude Code PreToolUse hook checks every command before it runs. Mine blocks any Salesforce deploy to production with exit code 2. The command never runs, and Claude sees why. https://t.co/CZzXc3WP7s
Free Salesforce certification retake: take an eligible exam by Nov 17, and if you don't pass, your retake is free between Jan 4 and Apr 7, 2027. You have to opt in first. https://t.co/EK5VwmrLYg
One file makes Claude Code follow your Salesforce rules every session. Here are the five I put in CLAUDE.md: bulkified triggers, no SOQL in loops, tests every time. https://t.co/fJCxerkPGd
Salesforce just retired its DX MCP server. Nothing breaks today, but it gets no more fixes. Here's where to move, and three checks before you switch your team. https://t.co/yUed8WklCy
Build once, ship to multiple interfaces. 🚀
With the open-source Pronto Sample App, explore the Headless Toolkit and learn how hosted MCP servers can securely expose Salesforce business logic, agents, and actions to third-party agents and apps like Slack, Claude and more.
Explore, deploy, and extend the app to put these patterns into practice. 👉 https://t.co/ovs8I0U2lD
@jitendrazaa Architects, honestly. The tooling speeds up build, but the judgment calls on data model, integration boundaries, and what the agent is allowed to do still sit with that role.
Broke a flow? Winter '27 lets you rewind it. Edit History shows every save and restores the good one, and element groups turn a 40-box flow into a few named sections. https://t.co/y1nZ6kcu2e
Agent Skills now lives under the Linux Foundation, next to MCP. The skills you write for your Salesforce org aren't locked to one tool anymore. One caution before you install any: https://t.co/o12x4e1lTC