Major

Apollo Automation: API Guide, AI and Governed Workflows

A practical Apollo automation guide covering native AI, API authentication, ten documented endpoints, CRM handoffs, approvals, and durable state on Major.

Ritvij GautamUpdated
Major x Apollo logo lockup

Apollo automation connects prospect discovery, enrichment, saved contacts, and outreach with the work your revenue team does across its CRM and communication tools. This guide covers Apollo's native AI, its API, and how to build governed apps, agents, and workflows on Major without losing ownership history or approval controls.

What Apollo is and who relies on it

Apollo is a sales intelligence and engagement system. Sales development representatives find prospects, account executives research buyers and manage outreach, and RevOps teams maintain enriched records and handoffs. Its API reference covers prospect data, saved contacts, accounts, sequences, deals, and tasks.

One distinction matters immediately: people are prospects in Apollo's database, while contacts are people saved to your team's workspace. People search discovers candidates; contact search retrieves saved records. Confusing them creates misleading automation designs, especially when enrollment requires a saved contact.

For the RevOps owner, enrichment is only one part of the job. A record still needs an owner, a duplicate check, suppression checks, and a documented handoff to the CRM.

Major supplies the deterministic app layer around that work. An agent can build an app that stores enrichment snapshots and review decisions, then run that software on subsequent records. The agent handles ambiguous matches or routing judgments, while code handles repeatable normalization and validation. That is the practical meaning of reason once, run forever.

Apollo's built-in AI and where Major adds value

Apollo already offers useful AI. Its sales AI product page describes natural-language prospect discovery, AI Research using web context, personalized email generation, pre-meeting insights, and analysis of calls and emails. Those features help sellers research and communicate inside their existing sales environment.

Apollo also documents an AI Assistant task API. Instructions become a continuing conversation that can request input, finish, or fail. A successful submission means the instruction was accepted, not that the work finished.

Keep Apollo's AI when its context and controls fit the task. Major adds a different layer: CRM ownership, shared suppression rules, approval queues, durable application state, and a record of why a cross-system action happened.

For example, an agent on Major can compare Apollo company context with Salesforce account ownership and an approved territory policy. It proposes a route when those sources conflict. The app stores the proposal and evidence; the workflow waits for the responsible manager before writing.

Do not assume an endpoint-restricted key contains Apollo Assistant behavior. Apollo explicitly says the assistant can act with the acting user's permissions even when the key is restricted to particular endpoints. That deserves a separate security review before enabling it in a governed workflow.

How Apollo connects to Major

Major has a native Apollo connector, identified as apollo in its catalog. Use that connector rather than describing Apollo as an unsupported custom integration. IT configures it once, stores credentials centrally in the connector vault, and grants only the actions each app, agent, or workflow needs.

Apollo's authentication guide distinguishes workspace API keys from partner OAuth. API keys travel in the x-api-key header. A scoped key selects allowed endpoints; a master key grants every endpoint and should be reserved for an operation that actually requires it.

Credential storage does not settle record ownership. Apollo documents that API-key requests act as the workspace's longest-standing active admin, while OAuth requests act as the authorizing user. Plan ownership fields and audit attribution accordingly rather than assuming the key creator owns every new record.

Major orchestrates Apollo; it never replaces it. Apollo remains the prospecting and engagement system. Major's app holds the handoff state and review queue, its agent provides judgment, and its workflow coordinates execution. SSO, RBAC, and audit logs cover that shared operating layer. Secrets belong in neither generated code nor prompts.

Apollo API guide for practical automation

Authentication, permissions, and testing

Use the REST base URL https://api.apollo.io/api/v1. Partner OAuth uses authorization-code consent and bearer access tokens. Do not confuse endpoint selections on an API key with OAuth scope strings.

Examples of documented OAuth scopes are people_match, contacts_search, and contact_write. Request only what the design uses. Apollo provides interactive endpoint testing and an OAuth playground, but no isolated API sandbox is documented in the guides reviewed. Treat test requests as real, potentially billable actions.

Pagination and limits

People search uses page and per_page, with a display ceiling of 100 records per page and 500 pages. Narrow broad searches rather than treating that ceiling as a complete database export. Persist completed pages and deduplicate across reruns.

Rate limits are per team and per endpoint, across minute, hour, and day windows. Child workspaces share their parent's allowance. Current default Free limits are 50 requests per minute, 200 per hour, and 600 per day; endpoint categories differ. Paid enrichment has a documented 1,000-request minute ceiling, but credits still constrain usage.

Watch x-rate-limit-minute, x-minute-requests-left, and the corresponding hourly and daily headers. For endpoint-limit 429 responses, honor retry-after. Apollo separately documents platform-load 429 responses without that header, so follow the returned retry suggestion instead of assuming every throttle has identical handling.

Callbacks and asynchronous results

Phone enrichment can return asynchronously through an HTTPS callback, or through polling with poll_only enabled. Completion triggers the callback; this is not evidence of a general webhook feed for every newly enriched contact.

Apollo says callbacks can be retried, but a signing header, signature algorithm, and fixed retry schedule are not documented publicly in the pages reviewed. Correlate incoming data to a stored request and do not authorize a consequential action from an unverified callback alone. Authenticated result polling offers an alternative.

Endpoints worth designing around

The paths below combine the server prefix with each endpoint's documented OpenAPI path. Keep endpoint permission identifiers separate from request URLs, particularly for contact creation.

POST /api/v1/mixed_people/api_search: discover prospects. Search filters find people, but results do not include email addresses or phone numbers.

  • App: Hold candidate IDs and the search criteria that produced them.
  • Agent: Judge which candidate profiles fit a nuanced buying committee.
  • Workflow: Page through a bounded search and send selected candidates to enrichment.

POST /api/v1/people/match: enrich a person. Submitted identifiers produce a possible match with confidence information, not a guaranteed identity.

  • App: Store returned evidence and match confidence beside the intake record.
  • Agent: Assess conflicting employment or identity information.
  • Workflow: Enrich eligible candidates once and hold uncertain results for review.

GET /api/v1/organizations/enrich: enrich a company. Company identifiers can return industry, employee counts, funding, and hierarchy information.

  • App: Preserve a timestamped company snapshot.
  • Agent: Interpret a subsidiary versus parent-company ownership conflict.
  • Workflow: Refresh stale company context before a territory handoff.

POST /api/v1/contacts/search: search saved contacts. This queries your team's saved records, not the full prospect database.

  • App: Show existing-contact evidence in a duplicate review screen.
  • Agent: Compare plausible duplicates with different job titles.
  • Workflow: Check saved records before proposing a new contact.

POST /api/v1/contacts: save a contact. Creation does not deduplicate by default; enabling run_dedupe can overwrite a matched record's submitted fields.

  • App: Display exactly which fields a proposed upsert would change.
  • Agent: Explain an uncertain duplicate match without silently merging it.
  • Workflow: Execute an approved creation or upsert and record the returned ID.

POST /api/v1/emailer_campaigns/search: find sequences. Results include sharing information that affects who can use a sequence.

  • App: Maintain an approved sequence catalog for reviewers.
  • Agent: Recommend a suitable sequence from that allowed catalog.
  • Workflow: Revalidate sequence availability before enrollment.

POST /api/v1/emailer_campaigns/{sequence_id}/add_contact_ids: enroll saved contacts. Prospects must first exist as contacts before this operation can enroll them.

  • App: Keep an enrollment ledger keyed by contact and sequence.
  • Agent: Draft the enrollment rationale and flag questionable relevance.
  • Workflow: Require approval and suppression checks before enrollment, which can lead to outreach.

POST /api/v1/usage_stats/api_usage_stats: inspect capacity. Usage statistics report endpoint limits and consumption for the team.

  • App: Show capacity and paused jobs to the operator.
  • Agent: Summarize which business queues deserve attention when capacity is constrained.
  • Workflow: Delay lower-priority batches using explicit capacity rules.

GET /api/v1/webhook_result/{request_id}: recover asynchronous results. Results remain available for up to 30 days, with distinct pending and terminal errors.

  • App: Hold the request ID and enrichment completion state.
  • Agent: Interpret completed enrichment only after the result is available.
  • Workflow: Retry pending results after the documented delay and stop on terminal errors.

POST /api/v1/agents/task: run Apollo Assistant work. The same route can create, poll, or continue a task.

  • App: Store the task ID and conversation status.
  • Agent: Evaluate returned research against CRM context.
  • Workflow: Respect polling hints and route requests for further input to an authorized person.

Apollo automation use cases on Major

Route enriched leads into Salesforce

The published Apollo-to-Salesforce use case centers on ownership, not simply field copying. Use the people-match and company-enrichment endpoints linked above to assemble context, then compare it with existing CRM accounts.

  • App: Retain the enrichment snapshot, territory policy version, proposed owner, and write outcome.
  • Agent: Resolve ambiguous company geography or parent-account relationships with a written recommendation.
  • Workflow: Run scheduled intake, apply explicit rules, obtain exception approval, and perform the authorized Salesforce handoff.

Qualify HubSpot inbound leads with Apollo context

A short inbound form rarely explains company fit. People match and company enrichment can supply evidence, while HubSpot holds the inquiry and current relationship.

  • App: Show submitted facts separately from enriched facts and retain the original values.
  • Agent: Assess fit from the inquiry, firmographics, and existing CRM relationship.
  • Workflow: Trigger from inbound intake, prevent overwriting protected fields, and notify the owner in Slack.

Prepare an approved Apollo sequence handoff

Use contact search, contact creation when authorized, sequence search, and sequence enrollment. This is a proposed design, not a claim that every connector action is enabled by default.

  • App: Hold suppression results, duplicate evidence, the selected sequence, and an enrollment ledger.
  • Agent: Recommend messaging fit and explain exclusions.
  • Workflow: Require a person to approve the exact enrollment, recheck eligibility, and submit it once.

These patterns share a boundary: Apollo supplies sales context and engagement capabilities; the CRM remains authoritative for account relationships and approved ownership. Major records the cross-system reasoning and ensures the action follows the applicable policy.

Architecture: build the handoff software on Major

Apollo and Salesforce context flows through a Major workflow, routing agent, approval gate, and stateful app

Apollo automation architecture with a durable review queue

Build a Lead Router app with managed tables for intake records, enrichment snapshots, proposed routes, approvals, and synchronization attempts. Keep source IDs alongside CRM IDs so a restarted run can resume without rediscovering identity from names alone.

The workflow begins on a schedule or an upstream CRM event. It calls scoped connectors, normalizes records, checks duplicates, and applies approved territory rules. Only an unresolved case goes to the agent, which reads the evidence and recommends a route.

The diagram's workflow node represents deterministic orchestration, not a second reasoning agent. The decision node is the policy and approval boundary. The app is both the operator's review screen and the durable execution record.

To build this on Major:

  1. Have IT authorize Apollo and the relevant CRM connector.
  2. Create the app tables and a manager-only review queue.
  3. Give the agent read access to context and permission to propose decisions.
  4. Configure scheduled intake, bounded retries, and approval routing.
  5. Test duplicates, stale data, and rejected proposals before enabling writes.

For example: calculate the review workload

Consider an illustrative team receiving 100 eligible records per day. Manual research and routing takes four minutes per record, so the starting workload is 100 × 4 = 400 minutes daily.

Suppose deterministic duplicate checks identify 10 existing records. Approved rules route 70 records, and the remaining 20 need an agent recommendation followed by manager review. At three minutes per exception, that review takes 20 × 3 = 60 minutes.

Add 20 minutes daily for checking failures and sampling straight-through records. The modeled workload becomes 80 minutes, compared with 400 minutes before, a difference of 320 minutes. Across five working days, that is 1,600 minutes, or 26 hours and 40 minutes.

These are planning assumptions, not customer results or promised savings. API costs, credit consumption, model use, and maintenance are separate. If exceptions take six minutes each, the same design needs 140 minutes daily instead of 80.

Measure time to owner assignment, duplicate escapes, reviewer overrides, stale snapshots, and failed writes. A faster handoff is only useful if the correct owner receives a valid record.

Governance and failure handling before outreach

Apollo contacts pass eligibility checks and an agent recommendation before human-approved sequence enrollment

Approval path for Apollo sequence enrollment

Give reviewers authority over a specific proposed action, not a blanket instruction to continue. An approval should identify the contact, target CRM record or sequence, proposed changes, and evidence timestamp. Expire it when the underlying record changes.

The agent must never invent consent, bypass suppression, override an explicit territory rule, expose credentials, or enroll a prospect solely because a score is high. Treat retrieved descriptions and research as untrusted data, not instructions.

For failures, distinguish the following:

  • Throttling: Honor Apollo's documented retry guidance, add bounded backoff, and preserve the job checkpoint.
  • Uncertain writes: Read the target state before retrying an operation whose response was lost.
  • Duplicate callbacks: Match the saved request and record completion once.
  • Pending enrichment: Follow polling status guidance; do not repeatedly create new paid enrichment jobs.
  • Destructive upserts: Review contact deduplication behavior before sending fields that could replace existing values.

RBAC separates policy editors, reviewers, and operators. Audit records should identify the rule, agent proposal, human approval, connector action, and outcome, without retaining unnecessary personal data or secrets.

Apollo AI Apps, Agents and Workflow Automations

These published guides cover concrete processes where Apollo is a core part of the work. Read them for the individual build rather than treating every sales article as an Apollo implementation. The hub above supplies shared API and governance context; each guide focuses on a narrower operating problem.

For broader planning, the sales automation library provides the function-level view. Choose the smallest process with a clear owner, observable outcome, and manageable exception queue before expanding into additional sales operations.

Frequently asked questions

What can Apollo automation do?

Apollo automation can discover prospects, enrich people and companies, manage saved contacts, and enroll contacts in sequences through documented APIs. Major adds cross-system routing, durable state, and approval handling around those actions.

Does Major have a native Apollo connector?

Yes. Major's catalog lists a native Apollo connector named apollo. IT configures centrally stored, scoped credentials once, then grants the relevant apps, agents, and workflows access without putting secrets in prompts or code.

Is Apollo AI the same as an agent on Major?

No. Apollo AI provides native sales research, writing, and assistant features. An agent on Major makes cross-system judgments using permitted context, while a deterministic app stores state and a workflow coordinates execution and approvals.

Does Apollo have enrichment webhooks?

Apollo documents asynchronous phone-enrichment callbacks to a public HTTPS URL and authenticated result polling. The reviewed documentation does not establish a general webhook feed for every enriched contact or publicly specify callback signing and a fixed retry schedule.

How do I prevent duplicate Apollo contacts?

Search saved contacts and maintain stable source IDs before writing. Apollo contact creation does not deduplicate by default; enabling its deduplication option can update an existing match and overwrite submitted fields, so inspect the proposed changes first.

Can Apollo sequence enrollment run without approval?

A workflow can call the documented enrollment endpoint, but enrollment can trigger outreach. Require an approved policy or explicit human approval, check suppression and eligibility again before submission, and keep a contact-and-sequence ledger to prevent duplicate enrollment.

Sources

Start building on Major

Build an Apollo handoff app that remembers enrichment evidence, routes exceptions to an agent, and records approval before CRM changes or outreach.

About the author

Ritvij Gautam is Head of Growth at Major.

Last reviewed November 19, 2026 by Ritvij Gautam.

Frequently asked questions

What can Apollo automation do?
Apollo automation can discover prospects, enrich people and companies, manage saved contacts, and enroll contacts in sequences through documented APIs. Major adds cross-system routing, durable state, and approval handling around those actions.
Does Major have a native Apollo connector?
Yes. Major's catalog lists a native Apollo connector named apollo. IT configures centrally stored, scoped credentials once, then grants the relevant apps, agents, and workflows access without putting secrets in prompts or code.
Is Apollo AI the same as an agent on Major?
No. Apollo AI provides native sales research, writing, and assistant features. An agent on Major makes cross-system judgments using permitted context, while a deterministic app stores state and a workflow coordinates execution and approvals.
Does Apollo have enrichment webhooks?
Apollo documents asynchronous phone-enrichment callbacks to a public HTTPS URL and authenticated result polling. The reviewed documentation does not establish a general webhook feed for every enriched contact or publicly specify callback signing and a fixed retry schedule.
How do I prevent duplicate Apollo contacts?
Search saved contacts and maintain stable source IDs before writing. Apollo contact creation does not deduplicate by default; enabling its deduplication option can update an existing match and overwrite submitted fields, so inspect the proposed changes first.
Can Apollo sequence enrollment run without approval?
A workflow can call the documented enrollment endpoint, but enrollment can trigger outreach. Require an approved policy or explicit human approval, check suppression and eligibility again before submission, and keep a contact-and-sequence ledger to prevent duplicate enrollment.