Salesloft Adds Native Claude & AI-Adoption Metrics

Salesloft’s June 2026 release does something most AI feature announcements skip: it adds a way to measure whether the AI is actually being used. Alongside native Claude access through the MCP Connectors Directory, the update introduces three new AI-usage metrics to Analytics — and that measurement layer, not the connectivity, is the part RevOps leaders should read twice.

The headline features are vendor-announced as part of a routine release (anticipated June 9, 2026): three new AI metrics in Analytics, a native connection from Salesloft into the Claude desktop app via MCP for customers on the Agentic add-on, and Cadence Collections — an organizational layer for bulk-managing many cadences at once. None of it is exotic in isolation. Together it sketches a quietly important pattern: ship the agents, connect them to the AI tools reps already live in, then prove the whole thing is getting used.

This guide covers exactly what shipped and what each capability does, where the three metrics surface (they are not all in the same report), how the Claude integration is gated, how Salesloft now sits against Outreach and the rest of the sales-engagement field, and what a budget-conscious RevOps team should actually do with it. Every figure below is sourced to vendor release notes or named primary sources; where a number is vendor-commissioned or third-party, we say so.

The June 2026 release notes (anticipated release date June 9, 2026) bundle three distinct capabilities. First, three new AI metrics in Analytics. Second, native Claude access through the MCP Connectors Directory in the Claude desktop app, available to customers on the Salesloft Agentic add-on. Third, Cadence Collections, an organizational layer that lets admins manage settings, permissions, and CRM-sync rules across many cadences at once. We frame all three as vendor-announced capabilities rather than independently verified outcomes.

This is a maintenance-cadence release, not a flagship launch — which is precisely why the adoption metrics are worth attention. They are not marketed as the centerpiece, but they answer a question most AI rollouts cannot: of the agents we turned on, which ones are people actually using? For a RevOps team trying to justify an AI budget, that is a more useful signal than another feature toggle.

One bit of context matters before going deeper. The agents whose usage these metrics track are not new. Salesloft introduced an Account Research Agent and a Person Research Agent in its “Closing Power” suite in late 2025, and added a batch of new AI agents in its Spring 2025 release. The June 2026 release does not ship a new wave of agents — it ships the instrumentation to see whether the agents already in the product are being adopted.

The single most useful operational detail in the release notes is also the easiest to miss: the three new metrics do not all surface in the same report. Getting this wrong means building a dashboard that quietly omits a metric. Per the release notes, Account researched and Person researched appear in both the Accounts Report and the Team Performance Report, while Agent tasks completed appears exclusively in the Individual Coaching Report.

That placement is itself a design statement. Putting Agent tasks completed in the Individual Coaching Report frames agent usage as a coaching conversation — something a manager reviews per rep, not a vanity number on a team dashboard. It treats “is this person using the agents we gave them?” the way good sales orgs treat call volume or email cadence: a leading indicator you coach against, not a score you celebrate.

The broader point is that “Agent tasks completed” is a different kind of metric than most AI dashboards report. The common proxy is “AI features enabled” — a setup statistic that says nothing about behavior. A completed-task count is closer to the real question. It is a usage signal, not an outcome signal: it tells you the agent ran, not that it created pipeline. But it is a meaningful step up from counting toggles, and it is rare to see it exposed natively in a sales-engagement platform’s analytics. If you are building the reporting layer around metrics like these, grounding agents in live CRM and pipeline data is the emerging standard for GTM stacks — a pattern we unpack in our piece on headless data layers for grounding AI agents.

The second headline feature: customers on the Salesloft Agentic add-on can now connect Salesloft directly to Claude through the MCP Connectors Directory in the Claude desktop app. The Model Context Protocol is the open standard that lets an AI assistant call into external systems for live data and actions; once connected, a rep can work with Salesloft data inside Claude without switching applications. If MCP tool-use is new to you, our guide to how Claude’s MCP tool-use works covers the mechanics.

This did not appear from nowhere. Clari + Salesloft launched its MCP Server on April 14, 2026 — opening the platform’s live data (pipeline movement, deal activity, customer interactions) to external AI tools including Claude, ChatGPT, Microsoft Copilot, Gemini, and Salesforce Agentforce. The June release is the in-Claude connector-directory experience layered on top of that earlier server. The April announcement also included full audit logging and governance controls, which is what makes the new adoption metrics a natural complement rather than a coincidence.

There is a strategic angle the vendor framing does not state plainly. Pairing the MCP server’s audit logging with the new in-Analytics adoption metrics is, in effect, a governance play: centralized observability of what the AI is doing, sitting next to centralized configuration in Cadence Collections. That combination — see what the agents do, and control how cadences behave from one place — matters because the market is deploying agents faster than it is governing them.

For context on why that governance gap is real: Anthropic’s Claude Connectors Directory has itself grown quickly since launching in mid-2025, with third-party counts in 2026 ranging from roughly 340 to 430-plus verified integrations depending on the source and date. The connectivity is proliferating; the measurement and governance to match it is the harder, less-shipped half.

The least glamorous feature in the release may be the one that saves the most hours. Cadence Collections introduce an organizational layer that lets admins manage settings, permissions, and CRM-sync rules across multiple cadences simultaneously. Cadences added to a collection inherit parent-level configuration, and visual indicators show when a setting is controlled at the Collection level rather than on the individual cadence.

Anyone who has administered Salesloft at scale knows the prior pain: settings, permissions, and sync rules lived per cadence, so changing a policy across dozens of cadences meant editing each one by hand. Collections turn that into inherited configuration. The one caveat worth flagging up front: existing standalone cadences require manual migration into a collection — this is not a retroactive, automatic grouping, so plan a migration pass rather than expecting your current cadences to organize themselves.

Salesloft is not first to Claude-over-MCP. Its chief competitor, Outreach, announced general availability of its own MCP server on February 24, 2026 — making it, at the time, the first revenue orchestration platform to connect to Claude via MCP. Outreach’s MCP requires an Amplify-enabled license and is read-only: it cannot create sequences or enroll contacts from Claude. Salesloft’s June connector lands later on the connectivity axis, but pairs it with the in-Analytics adoption metrics that Outreach has not published.

The competitive read is straightforward. On pure connectivity, Outreach moved roughly four months earlier (February 24 to June 9 is about 105 days). But connectivity is becoming table stakes; both platforms can put their data inside Claude. The place Salesloft differentiates in this release is the measurement layer — exposing agent and research usage natively in Analytics is the half of the problem most of the field, Outreach included, has not yet shipped.

Here is the argument that makes this release more than a changelog entry. The hard problem in enterprise AI right now is not adding agents — vendors are shipping them quickly. The hard problem is knowing which agents your team actually uses, and whether that usage is producing anything. A metric like “Agent tasks completed” is one of the first native attempts inside a sales-engagement platform to answer the usage half of that question directly, rather than inferring it from a feature-enabled flag.

Looking forward, expect this to become a category expectation rather than a Salesloft quirk. Once one platform makes “is the AI actually being used?” a first-class report, buyers will start asking competitors for the same view, and the marketing story shifts from “how many agents we have” to “how much of your team is using them.” The vendors that win the next phase of GTM tooling will likely be the ones that can prove adoption and govern it, not just the ones with the longest agent roster. For a wider map of where these agents fit, our buyer’s guide to the AI SDR agent landscape sets the broader context for where sales-engagement platforms sit.

Each capability has a different entitlement. The metrics are the most broadly available; the Claude connection is the most gated. Here is who can use what, per the release notes.

For budget-conscious RevOps buyers comparing Salesloft against Outreach, the access model is part of the decision. Salesloft’s Claude path sits behind a separate Agentic add-on SKU; Outreach’s read-only MCP requires its Amplify-enabled license but, per its GA announcement, is not framed as a separate add-on SKU. Neither is free, and the capabilities differ — read-only versus read-write, metrics versus none — so the right comparison is total cost against the specific workflow you need, not a feature checkbox.

The practical decision tree here is short, because the release does not change strategy — it changes what you can measure and how you administer at scale. Map your move to the situation you are in.

If you want help wiring these adoption metrics into a reporting layer your leadership will actually read — or evaluating whether the Claude integration earns its add-on cost against your specific workflow — that is exactly the kind of scoping our CRM & automation engagements handle. For the broader question of measuring and governing AI across a GTM stack, our AI digital transformation work starts with exactly this kind of adoption-and-governance assessment.