Major

Salesloft AI Agents: API Guide and Governed Automation

Understand Salesloft's native AI agents, nine verified API operations, and how Major adds cross-system review apps, scoped agent judgment, and approval workflows.

Ritvij GautamUpdated
sales team reviewing pipeline on laptop (Salesloft AI agents and automations)

Salesloft AI agents research buyers, prioritize seller work, and assist with emails and deal reviews inside Salesloft. This guide explains those native features, the API behind Salesloft automation, and how Major adds governed cross-system apps, agents, and workflows for RevOps teams.

What Salesloft is and who relies on it

Salesloft organizes revenue execution around buyer engagement, conversations, deals, and prioritized seller actions. Its platform overview describes Cadence for multichannel engagement, Conversation Intelligence for call insights, Deals for opportunity visibility, and Rhythm for prioritizing daily work from buyer signals.

Sales development representatives use cadences to organize prospecting. Account executives work opportunities and prepare for buyer conversations. Sales managers inspect pipeline risk and coach sellers, while RevOps maintains record relationships, operating standards, and reporting across the sales stack.

For RevOps, the difficult work often sits between systems: linking a Salesloft person to the correct CRM opportunity, distinguishing an engagement gap from a broken sync, and preserving a manager's previous decision. More generated text does not solve those problems by itself.

Major is the enterprise platform where agents build the software they run on. For Salesloft operations, that software can be a deterministic review app with matching logic, managed database tables, evidence views, and approval history. An agent supplies judgment only after the repeatable checks produce a case worth reviewing.

Salesloft AI agents: native capabilities and the Major layer

Salesloft's native AI agents page names capabilities across prospecting, closing, retention, and coaching. Account Research and Person Research summarize account and buyer context. Buyer Identification recommends stakeholders, while Ask Salesloft lets sellers ask questions about an account.

Rhythm Prioritizer organizes daily actions. AI Email Steps for Cadence creates personalized emails when steps run, and the Generative AI Email Assistant composes or edits messages. Deal Summaries provides opportunity overviews, Sales Methodology Extraction pulls call information into frameworks, and Stalled Deals Insight surfaces stalled deals in Command Center.

These are meaningful native capabilities. A cross-system engagement review should not be described as something Salesloft cannot do at all. The question is whether your policy also requires Salesforce opportunity context, Gong evidence, a durable exception register, or a specific approval record outside the seller workspace.

Major adds that operating layer. The app stores evidence and decisions, the agent judges ambiguous cases, and the workflow routes approved actions. Native Salesloft AI remains useful for seller assistance. Public product descriptions do not establish API access to every named AI feature, so the API examples below use only documented resources.

How Salesloft connects to Major

Salesloft has no native or Pipedream connector in the catalog: it connects as a custom API connector, with base URL and authentication configured once by IT. Configure the origin as https://api.salesloft.com, identified in the official reference site's configuration, and use documented versioned resource paths such as the People API.

For a customer-owned integration, IT can provision a scoped Salesloft API key that acts on behalf of its issuing user. Partner integrations require a registered OAuth application. Keep secrets in Major's central connector vault, never in generated code, browser fields, or agent prompts.

Start with read access and test record visibility before enabling writes. The agent receives permission to call selected connector operations, not unrestricted access to every capability the underlying credential permits. SSO, RBAC, and audit logs cover the app, agent, and workflow.

Major orchestrates Salesloft; it never replaces it. Salesloft remains the engagement workspace. Major holds the custom review state and cross-system policy, including when a proposed action must wait for a person.

Salesloft API guide for governed automation

Authentication, scopes, and test options

Customer API keys and OAuth access tokens use Bearer authorization. Authorization-code OAuth returns refresh tokens that rotate when refreshed. The private client-credentials flow is admin-enabled, assumes its creator's permissions, and requires obtaining a new access token rather than using a refresh token.

The scope guide lists people:read, accounts:read, cadences:read, calls:read, and emails:read for relevant read operations. Email bodies and subjects require the privileged scope email_contents:read. Verify each operation's authorization requirements before adding write access.

The OAuth guide documents Test Authorization for checking an app's authorization flow. A separate public API sandbox is not documented in the official pages reviewed here. Use isolated test records in an approved environment, with outbound actions disabled.

Pagination, limits, and events

Pagination uses page and per_page, generally defaulting to 25 records with a maximum of 100, subject to each endpoint. Response metadata includes the next page. Rate limits are shared team-wide at 600 cost points per minute by default, not 600 calls for each integration. Deep pages cost more. Watch x-ratelimit-endpoint-cost and x-ratelimit-remaining-minute.

Webhooks deliver HTTP POST notifications and retry failures three additional times, 15 seconds apart. Event types include person updates, cadence membership changes, call creation, and email updates. Delivery headers document x-salesloft-signature: verify its HMAC-SHA1 hex digest against the raw body using the subscription's callback_token as the key.

Read people: GET /v2/people

Retrieves person records with documented filtering, paging, and sorting.

  • App: Maintain the person-to-CRM matching register.
  • Agent: Assess an ambiguous contact association from permitted context.
  • Workflow: Refresh known people before a daily engagement scan.

Read accounts: GET /v2/accounts

Retrieves account records for company-level context.

  • App: Group exceptions by account rather than duplicating company details.
  • Agent: Interpret whether several quiet contacts indicate broader account risk.
  • Workflow: Refresh account context before assembling a manager review batch.

Read cadences: GET /v2/cadences

Retrieves cadence definitions visible to the integration.

  • App: Display the cadence behind each flagged engagement gap.
  • Agent: Consider cadence purpose when recommending a follow-up.
  • Workflow: Refresh the approved cadence catalog on schedule.

Read memberships: GET /v2/cadence_memberships

Retrieves mutable relationships between people and their cadence history.

  • App: Preserve membership snapshots and review history separately.
  • Agent: Distinguish intentional sequencing from an unexplained engagement gap.
  • Workflow: Check membership context before applying inactivity rules.

Read emails: GET /v2/activities/emails

Retrieves logged email activity records.

  • App: Show email timestamps in the evidence timeline.
  • Agent: Weigh available email context against the proposed next step.
  • Workflow: Combine email and call timestamps to calculate last logged touch.

Read calls: GET /v2/activities/calls

Retrieves logged call records, not a promise of unrestricted transcript access.

  • App: Show call activity alongside email evidence.
  • Agent: Judge whether the permitted call context supports escalation.
  • Workflow: Reconcile new calls before marking an opportunity quiet.

Enroll a person: POST /v2/cadence_memberships

Adds a visible person to a visible cadence, subject to user and cadence permissions.

  • App: Present the exact person, cadence, and acting seller for approval.
  • Agent: Recommend an appropriate approved cadence with supporting rationale.
  • Workflow: Revalidate permission and membership before executing approved enrollment.

Record a note: POST /v2/notes

Creates a Salesloft note.

  • App: Store the proposed note and returned record reference.
  • Agent: Draft an evidence-based internal handoff summary.
  • Workflow: Create the note only after the configured content approval.

Subscribe to events: POST /v2/webhook_subscriptions

Creates a subscription using valid authentication and event-dependent scopes.

  • App: Track subscription ownership and receiver health.
  • Agent: Assess exceptions assembled from relevant event context.
  • Workflow: Provision approved subscriptions and route verified receipts into processing.

Salesloft automation use cases to build on Major

Quiet opportunity review with Salesforce

Catch opportunities whose linked Salesloft activity has gone quiet without automatically declaring them lost.

  • App: Hold opportunity links, last-touch evidence, and owner review status.
  • Agent: Recommend re-engagement or escalation using stage and engagement context.
  • Workflow: Schedule the scan, validate joins, and route review items. Use the linked People, Cadences, Memberships, Emails, and Calls read endpoints above.

Gong coaching into a seller handoff

Turn a flagged Gong conversation into a proposed talk track rather than silently changing a shared cadence.

  • App: Preserve the coaching moment, source reference, draft, and approval.
  • Agent: Identify the objection and draft a suitable next conversation approach.
  • Workflow: Gather Gong context, use the People and Cadences reads, and create an approved handoff through the Notes write above. Cadence editing requires separately verified API support.

CRM-qualified enrollment review

Use Salesforce qualification context to propose Salesloft enrollment, with Slack notifying the responsible reviewer.

  • App: Display account matches, selected cadence, and approval status.
  • Agent: Recommend the cadence and explain exceptions to the usual playbook.
  • Workflow: Read People, Accounts, and Memberships, then use the enrollment write only after approval and a fresh eligibility check. Keep prospect-facing execution in Salesloft.

These are proposed operating patterns, not claims that installing a connector creates finished automations. Begin with a bounded queue and one owner, then add writes only when the review process works.

Architecture: build the Salesloft review layer on Major

Salesloft and Salesforce feed a Major workflow, judgment agent, approval gate, and stateful review app

Figure 1. Keep matching and state in software, with judgment in the agent.

Build the review layer as reusable software rather than asking an agent to scan the whole pipeline from scratch every morning. The same app gives managers a control surface and gives the workflow a durable execution record.

  1. Connect Salesloft through the custom API connector and attach the separately authorized Salesforce connector.
  2. Create app tables for record links, activity evidence, review items, and action attempts.
  3. Use a scheduled workflow to retrieve records, validate joins, calculate inactivity, and upsert exceptions.
  4. Give the agent scoped context for exceptions, including evidence freshness and prior review outcomes.
  5. Route recommendations into the app queue, then require approval for enrollment or record changes.
  6. Persist external record references and execution outcomes so interrupted runs can resume.

The app does not decide whether a buyer needs a different approach. The agent makes that judgment. Fixed rules handle matching prerequisites, permitted actions, and approval requirements. Once implemented, those repeatable checks run in code rather than spending model context on every record. Reason once. Run forever.

A worked example with explicit assumptions

For example, assume a sales manager reviews 200 open opportunities each week. Checking linked contacts, reading touch history, and writing a short follow-up assessment takes six minutes per opportunity. That is 200 × 6 = 1,200 minutes, or 20 hours weekly.

An illustrative Major implementation performs the record joins and inactivity calculations in deterministic code. Suppose 40 opportunities require judgment and review. The agent drafts a rationale for each, and the manager spends three minutes validating each recommendation: 40 × 3 = 120 minutes.

Allow another 60 minutes weekly for broken links, access exceptions, and checking scan completeness. The modeled operating time becomes 180 minutes, or three hours, compared with 20 hours before. The difference is 17 hours under these assumptions, not a measured customer result or a promise.

Measure the real baseline before launch. Track false flags, unresolved matches, approval time, and recommendations overturned by reviewers. If incomplete activity data sends too many records into the exception queue, repair the ingestion and matching logic before expanding the agent's remit.

Governance, approval, and failure handling

Verified Salesloft events reach a Major agent, then internal flags proceed while enrollment waits for human approval

Figure 2. Separate read-only detection from approved external changes.

Require a reviewer before enrolling a person, publishing a handoff note, changing CRM fields, or initiating prospect outreach. The agent must never treat its own recommendation as approval, infer consent from engagement, or resolve an uncertain identity match by guessing.

Verify webhook signatures before processing, persist receipt state, and deduplicate repeated payloads. The documented retry policy makes duplicate delivery a normal condition to design for. Use stable resource identifiers plus event context to distinguish a repeat delivery from a real update.

Throttle calls against the shared rate budget. Apply bounded backoff with jitter for transient failures. After an ambiguous write timeout, reconcile whether the action happened before retrying. Missing data must produce an incomplete status, not a false inactivity judgment.

Record the input evidence, agent rationale, policy version, approver, attempted action, and resulting external reference. Scope rep views to assigned records and manager views to permitted teams. Keep sensitive email content out of the agent context unless explicitly authorized and necessary.

Salesloft AI Apps, Agents and Workflow Automations

Every published Major use case that works with Salesloft, newest first. Each one names the Salesloft API objects involved and shows the app, agent and workflow split.

Frequently asked questions

What are Salesloft AI agents?

Salesloft AI agents are native capabilities for account research, buyer context, prioritization, email assistance, deal summaries, and coaching. Major can add a governed cross-system operating layer, with deterministic apps holding state and agents judging exceptions.

Does Major have a native Salesloft connector?

No. The current catalog connects Salesloft through a custom API connector. IT configures the base URL and scoped authentication once, with credentials kept in Major's central vault rather than code or prompts.

Can an agent automatically enroll people in Salesloft cadences?

The documented cadence membership API supports enrollment subject to visibility and permissions. In the proposed Major design, an agent recommends enrollment, a person approves it, and a workflow rechecks eligibility before executing the change.

Does Salesloft provide webhooks?

Yes. Salesloft documents subscriptions for events including person updates, call creation, email updates, and cadence membership changes. Verify the documented HMAC signature and handle duplicate delivery, including the three additional retries spaced 15 seconds apart.

What is the Salesloft API rate limit?

The documented default is 600 cost points per minute, shared across the team and its integrations. Endpoint cost and deep pagination affect consumption. Read the rate-limit response headers rather than assuming each request always costs one point.

Do Salesloft AI agents replace pipeline review?

Native features help with research, prioritization, and deal insight. A custom Major review process adds your cross-system evidence, approval policy, and durable decision history. Managers still approve consequential actions, and missing or conflicting evidence should return to a person.

Sources

These official pages support the product descriptions, authentication rules, and resource boundaries above. API availability and seller permissions should still be checked against the connected account before implementation.

Start building on Major

Build a Salesloft engagement review app with scoped connectors, evidence-backed agent recommendations, and human approval before consequential actions.

About the author

Ritvij Gautam is Head of Growth at Major.

Last reviewed November 20, 2026 by Ritvij Gautam.

Frequently asked questions

What are Salesloft AI agents?
Salesloft AI agents are native capabilities for account research, buyer context, prioritization, email assistance, deal summaries, and coaching. Major can add a governed cross-system operating layer, with deterministic apps holding state and agents judging exceptions.
Does Major have a native Salesloft connector?
No. The current catalog connects Salesloft through a custom API connector. IT configures the base URL and scoped authentication once, with credentials kept in Major's central vault rather than code or prompts.
Can an agent automatically enroll people in Salesloft cadences?
The documented cadence membership API supports enrollment subject to visibility and permissions. In the proposed Major design, an agent recommends enrollment, a person approves it, and a workflow rechecks eligibility before executing the change.
Does Salesloft provide webhooks?
Yes. Salesloft documents subscriptions for events including person updates, call creation, email updates, and cadence membership changes. Verify the documented HMAC signature and handle duplicate delivery, including the three additional retries spaced 15 seconds apart.
What is the Salesloft API rate limit?
The documented default is 600 cost points per minute, shared across the team and its integrations. Endpoint cost and deep pagination affect consumption. Read the rate-limit response headers rather than assuming each request always costs one point.
Do Salesloft AI agents replace pipeline review?
Native features help with research, prioritization, and deal insight. A custom Major review process adds your cross-system evidence, approval policy, and durable decision history. Managers still approve consequential actions, and missing or conflicting evidence should return to a person.