Home / Integrations / Anthropic
Route reasoning-heavy steps to Claude inside your Agentforce flow
Anthropic builds the Claude family of language models, designed for careful reasoning and steerable, safety-conscious behaviour. Agentforce runs Salesforce's own models by default, but some teams standardise on Claude for the steps where reasoning quality and predictable behaviour matter most. Emerge Digital designs that integration where you want it and where Salesforce supports connected or bring-your-own models, so an agent can hand a specific task to Claude while orchestration, records, and governance stay on Agentforce. Emerge works closely with Claude day to day, so we scope those steps from experience rather than guesswork.
What this unlocks
- A Claude model available for defined reasoning steps — multi-step analysis, careful summarisation, or a judgement call that benefits from steerable behaviour — configured where Salesforce supports a connected model.
- Claude's instruction-following makes it easier to constrain a step to a narrow, reviewable task, so an agent's reasoning stays inside the boundaries you set.
- You decide which steps route to Claude and which remain on Salesforce's built-in models, with that split fixed in configuration rather than left to chance.
- Context sent to Claude and the outputs it returns pass through governance you own, so a Claude-backed step only sees what its permissions allow.
In the customer journey
A careful reasoning step under review
When a case needs multi-step analysis before an agent responds, Emerge routes that step to Claude, and the reasoning it produces returns to the Agentforce flow where a person confirms it before anything is acted on.
Summaries a team can trust
For condensing a long thread or history into a brief, faithful summary, the agent hands the step to Claude, and the result feeds back into Salesforce as context the next human or agent picks up from.
Claude tends to earn its place at the Engage and Optimize stages, where reasoning and faithful summarisation shape both the conversation and what the team learns from it afterward. It attaches to specific steps rather than the whole journey — Agentforce still orchestrates, and Salesforce stays the system of record. The value is depth on the moments where judgement matters.
How Emerge integrates Anthropic
Emerge Digital connects Anthropic's Claude with Salesforce Agentforce as a consulting engagement, not a downloadable app. We work within what Salesforce supports for connected or bring-your-own models, configure the reasoning steps that route to Claude, and set the rules for what those steps may see. We do not claim Claude replaces Agentforce or that it slots in with a single toggle — we design where Claude fits, govern the data it receives, and keep a person in the loop wherever a decision carries weight.
How we structure an engagementRelated integrations
FAQ
Can we standardise entirely on Claude instead of Agentforce's models?
Not entirely — Agentforce runs Salesforce's own models through the Einstein Trust Layer. Where Salesforce supports connected models, Emerge configures Claude for the reasoning steps you choose, so it works alongside Agentforce rather than replacing it.
Do you just flip a switch to add Claude?
No. This is a services engagement. Emerge designs which steps route to Claude within Salesforce's supported approach, builds the configuration, and governs access — there is no self-install connector that wires Claude into Agentforce for you.
Why does Emerge working with Claude matter here?
Because scoping a Claude step well takes familiarity with how the models behave. Emerge uses Claude in its own work, so we tune prompts and boundaries from experience — while still governing access by permissions and keeping humans in the loop where stakes warrant it.
Ground your agents in Anthropic.
Tell us what your agents need to read and write in Anthropic, and we'll design the integration and the governance around it.
Talk to the practicePrefer email? Write to the practice instead.