v1.4.2: now in Codex, and in any chat

Now in Codex, and in any chat. The same 20 skills install as a Codex plugin from this repo,
and a start-here prompt runs the method in ChatGPT or any other chat. Everything sits beside
Apollo's own tools: where Apollo's AI actions fit a step, the operator names them as an option and
asks before running one.

• Codex plugin. codex plugin marketplace add creatop-gtm/apollo-operator, then
  codex plugin add apollo-operator@creatop. It lives in plugins/apollo-operator/ beside the
  untouched Claude plugin, rendered from the same source. Each skill also ships a version for a
  chat without a terminal, with the same cost gate, bounded reads, and deliverability rules.
  Installed from GitHub in a clean profile: 20 skills.
• Apollo's AI actions are offered, not avoided. If discovery lists an apollo_agent_* action
  for the job, the operator names it alongside the method, says what it may spend or create, and
  waits for a yes. What it returns still goes through List and Infrastructure before anything sends.
• Two broken references fixed. business-brief and apollo-icp-builder pointed to
  references/outbound-principles.md in their own folders; the file lives in
  operator-context/references/. Both layouts corrected.
• README and plugin description reworded to credit what Apollo's MCP and plugin do and to
  describe the library as sitting beside them.

v1.4.1: built to sit beside Apollo's own tools

Apollo published MCP v2 on 2026-09-23, a four-tool router in front of the full catalog, and its Claude Code plugin gained a strategy skill. This release re-measures Apollo Operator against the live endpoints and repositions it around one rule: where Apollo now does a step well, the library offers Apollo's way first and picks up where it hands off. Complementary, not parallel. Same 20 skills.

Kept current

• The router is the default. On the same account, v1 lists 85 tools and about 485 KB of schema; v2 exposes the same catalog through four tools and about 6 KB. The same search returns the same 7,829 people on both.
• The catalog grew by 11 tools in eleven days and v2 stayed at four, which is the design doing its job. Record collections, fields, dynamic AI enrichment, data sources, and CSV exports now dispatch through v2 and are in Apollo's docs.
• Reachability re-verified for the agent tools and the master-key path, and the v2 dispatcher calling convention documented with an example.

Positioned beside Apollo's GTM Strategist

Apollo's plugin now ships /apollo:gtm-strategist, which turns a pipeline goal into one recommended play with real reasoning discipline: diagnose before searching, evidence labelled verified or inferred, a small validation cohort, the tool's own approval gates. It runs on the ordinary connector and it is good.

Apollo Operator now offers it where Targeting starts and takes the handoff it describes itself: warmup and deliverability, independent verification, list composition, suppression, per-company caps, the attrition between a sized audience and a sendable list, and reporting past the reply. Where the two overlap, the operator chooses and the library says the option out loud.

Install from any directory: /plugin marketplace add creatop-gtm/apollo-operator, then /plugin install apollo-operator@creatop. Full detail in the CHANGELOG.

v1.4.0: MCP v2, Apollo's agent, and three new skills

Apollo shipped a four-tool MCP router, opened the MCP to master API keys, and put its own AI agent behind nine MCP tools. This release tests all three on a live account and rewrites the lane model around what held. Three new skills bring the library to 20.

MCP v2: prefer the router

https://mcp.apollo.io/mcp-v2 exposes four tools instead of 74: 5.6 KB of schema instead of 344 KB, every v1 tool still dispatchable, identical search results. The hand-maintained tool map is now a convenience and apollo_find_tools is the authority. The dispatcher is a free cost signal: anything routed through apollo_write_destructive spends credits or changes live state, so it gets a cost quote and a human first.

A master API key runs the MCP unattended, with no consent prompt and no session expiry. It also reaches tools the router documents but refuses to dispatch (record collections, CSV exports, dynamic AI enrichment, domain diagnosis), which we reported to Apollo.

Apollo's own agent: complement, do not compete

Nine apollo_agent_* tools are nine doors to one agent. We tested it. Planning is free, and it is good: it read the Context Center, planned four campaign concepts, sized four audiences live, and stopped for confirmation.

It is strong at the front of the motion and absent from the back: no deliverability check, no independent verification, no composition grading, no suppression, no warmup clock. That is the half this library exists for. Where the two overlap on targeting, planning, or copy, the skills now say the option out loud and let you choose. Whatever the agent returns still goes through List and Infrastructure before anything sends. Product how-to questions route to Apollo's own knowledge, not ours.

Reachable today only on the v1 endpoint with a master key. Check apollo_find_tools rather than trusting the changelog.

Three new skills

• campaign-reporting: results per sequence, step, variant, mailbox, seniority, and day of week, plus meetings booked and opportunities. Read-only, verified against the Apollo UI.
• apollo-account-audit: 14 read-only checks on an existing account, graded, stops at findings. Its first run found an active sequence sending nothing because its touches were never approved.
• account-based-outbound: when to reach several people at one company, how many, and which account tools carry it. Tested live.

Also new

• A compliance reference: opt-outs, the workspace GDPR setting, do-not-call screening, recorded-call consent. Operating rules, not legal advice.
• Install as a plugin from any directory: /plugin marketplace add creatop-gtm/apollo-operator, then /plugin install apollo-operator@creatop. About 1,400 always-on tokens per session.
• Company search, measured: 1 credit per call including the empty page, and a company row already carries founded year, revenue, NAICS, SIC, and headcount growth over 6, 12, and 24 months, so growth and sector grade off one free join. Plus a zero-credit preview sample for a prospect or client.
• Two targeting traps with numbers: a founder title filter returns Founding Engineers and Designers too (9 to 14% of a raw list), and "recently funded" includes acquisitions (34% of one universe).

Corrected with dates

• Sequences are readable now. apollo_emailer_campaigns_show returns every step and body; v1.3 said no lane could.
• MCP can stop a live sequence, with a trap: the update is declarative, so any step you omit is deleted. The CLI's sequences abort stays the kill switch.
• Step 3 is the shortest email after all, one or two sentences. v1.3 had reversed it.
• Single contact create matches on email and overwrites the existing record. Bulk create never updates on a match. Never test with a real address.
• AI-written touches are silently dropped on create. Read every touch back.

Full detail, with dates and measurements, in the CHANGELOG.

v1.3.0 — Call intelligence, credit control, and verify-first lists

Three new capabilities, all of which make a campaign cheaper to build and more likely to land.

Call intelligence feeds the brief

If your team records calls in Apollo, the best research material for your positioning is already in your account, and reading it is free.

conversations show returns a structured call_summary: objections, pain points, outcome, next steps, and pricing discussion. Five direct inputs to the brief, in your buyers' own words rather than marketing's.

A website is how a business describes itself. A recorded call is how its buyers describe their problem. When the two disagree, the call wins, and business-brief now says so.

Credit control

Credits are eleven separate pools, not one balance, and the ratio changes how you plan: the AI pool is 200× the contact-data pool. Apollo prices generation as near-free and data as scarce. So the expensive part of a campaign is finding and verifying people, not writing to them.

Three measured rules that cut real spend:

• The free has_email filter. Search returns it on every row, before you spend anything. Records without it come back empty or with a guessed address, and both still cost a credit. Filtering first took an enrichment yield to 99.7% verified.
• You pay per record attempted, not per email found. Misses cost the same as hits.
• Re-enrichment recharges. There is no "already paid for this one" discount, which is why suppression earns its place.

Plus the hard cap nobody documents: 10 records per enrichment call, and the CLI does not batch for you.

Verify-first lists

The pipeline is reordered: search → filter → dedupe → suppress → enrich → verify → write copy.

Verification moved ahead of copy because personalising an address that will never receive anything is pure waste. The evidence: 997 addresses graded verified by the provider were 564 safe to send, 413 catch-all, and 3 that would have bounced. Three bounces in a small first batch is a live deliverability incident on a young domain, and no amount of good copy prevents it.

Also new

• First-party intent as a fourth kind of angle, plus a table mapping every advanced filter to the kind of angle it builds.
• Bounce Guard documented, including the trap: it does not evaluate below 200 emails, so small sends and the first days of a ramp are unprotected.
• Follow-up craft: how the greeting tightens across a thread, and the four things to cut from every follow-up.
• Sequence build mechanics: the exact payload shape, step ids on update, and the schedule fields that decide when anything actually sends.
• A list gets into Apollo as a real list, in checkpointed batches, with the list attached properly.

Kept current

Five things moved under us since v1.2 and are now corrected with dates, so you can tell what changed and when:

• The flagship MCP-only filter example moves to NAICS, since --department-headcount landed in CLI v2.1.0. Eight of nine remain CLI-less and a new MCP-only family appeared, so the four-lane model got stronger.
• Per-contact personalisation no longer needs REST. apollo fields create exists now.
• conversations is a full CLI command group, not REST-only.
• Step 2 is the lightest email, not step 3.
• Pagination overlap is no longer guaranteed: zero duplicates across 43 pages, against 9% in July.

Read these before your next launch

• Bulk contact creation does not deduplicate, despite documenting that it does. Never retry on a timeout.
• Enrollment can fail silently, returning exit 0 with an empty campaign. Assert on the response body.
• Sequences cannot be read back. A human in the UI is the only verifier, so every skill that creates one now hands you the link.
• The MCP session can expire mid-task while the CLI token keeps working.

Still 17 skills. Built by Creatop, and everything above was found by running it, not by reading the docs.

v1.2.0 — Four lanes, named phases, re-engagement

Everything in this release came from running the library end to end on a live Apollo account rather than reading it. The previous audit was a full read of every skill and caught none of these. 17 skills.

The headline: there are four lanes, not three

apollo auth login stores an OAuth token, and that token authenticates against the same REST endpoint the MCP uses. So you get the complete MCP filter surface (NAICS, tenure, headcount growth, department headcounts) while writing results straight to disk — no API key needed. Verified against MCP on an identical query: same total, 3 MB to disk, nothing through the context window.

Lane	Auth	Results go	Best for
MCP	OAuth	Through context	Guardrailed, human-gated steps
CLI binary	apollo auth login	To disk	Bulk work, common filters
REST + the CLI's own token	Reuses the CLI's token	To disk	Bulk work needing advanced filters
REST + API key	API key	Anywhere	Building a separate service

Neither surface contains the other. Record collections, mailbox purchasing, and the Context Center are MCP-only. Deals, custom-field creation, rate-limit stats, and stopping a live sequence are REST/CLI only.

The rule it collapses to: search wide off-context, act narrow on MCP.

Added

• re-engagement — going back to the 90%+ of any list who finished a sequence and never replied. When it is safe (60-90 days), the permanent exclusions, and why "just following up" is not a reason.
• Operator voice — the communication standard, as prime directive 7. How the operator introduces itself, five named response shapes, a jargon gloss table, and ceremony matched to stakes. Written for someone who has never run outbound.
• The 85% rule — if a step would spend more than 85% of your remaining credits, it stops and offers you named choices instead of quoting a number.
• A required explanation gate before enrichment. Never hand over a free-stage list without saying what is missing and why.
• angles.md — running several angles against one ICP. Timing signals decay, technographic fit does not, structural facts never do.
• cli-recipes.md — verified commands, response-shape gotchas, credit table.
• Waterfall enrichment, phone reveal, Context Center, mailbox purchasing, per-contact merge personalization, website-visitor signals.

Changed

• Numbered levels replaced with named phases. Foundations → Context → Targeting → Infrastructure → List → Message → Launch → Iterate. Infrastructure moved ahead of List and Message because warmup is a 14-21 day wall-clock wait that must run in parallel.
• apollo-cli renamed apollo-operator, since Apollo now ships its own official skill under the old name. We reference theirs for command detail rather than duplicating it.
• The context-cost claim is now measured, and the obvious version was wrong: CLI and MCP return byte-identical payloads. The saving is preloaded tool schemas plus the ability to bypass context entirely.

Fixed

• Title filters silently over-match. founder also returns Founding Engineer, Founding Designer, Founding GTM — 9 to 14% of a raw list.
• Pagination overlaps. 3,238 rows contained 2,933 unique people. Per-company capping does not remove them, and you pay per duplicate.
• A data provider's catch-all flag is not a verification. One list reported clean on all 846 addresses; an independent verifier found 39% catch-all.
• MCP is the only lane that cannot stop a live send. Set up the CLI before you launch.
• people search -f csv is broken — the whole array lands in one cell.
• companies search splits results across .accounts and .organizations, ratio shifting by page.
• People search is free; company search costs 1 credit per call.
• Suppression now covers your other lists, not just the current one.

Full detail in CHANGELOG.md.

Built by Creatop.

v1.1.0 — CLI awareness + business brief + sending infrastructure

Apollo shipped a CLI and formalized three ways to run headless (MCP, CLI, raw API). This release makes Apollo Operator aware of all three, and adds the two skills that were missing from the front and back of the motion: understanding the business before targeting, and building the sending stack before sending. 16 skills total.

Added

• apollo-cli (Level 0, Access): the three lanes to run Apollo headless, the MCP-vs-CLI tradeoff, when Operator reaches for each, CLI setup, and a zero-credit auth-health preflight. The library stays MCP-native by default; CLI-backed search and enrich variants are planned for a future release.
• business-brief (Level 0.5): one readable brief.md capturing the business (identity, offer, buyers, pains, objections, proof, voice), the narrative context layer every other skill reads.
• sending-infrastructure (Level 4, Setup): build the sending stack from zero (dedicated domains, mailboxes, SPF/DKIM/DMARC/MX, redirect, warmup) plus the volume math for mailboxes and domains per daily send target.

Changed

• apollo-multichannel: reframed human-in-the-loop as the LinkedIn safety advantage, and added pacing (keep connection requests under 30/day, 20 to be safe, trickled across days).
• operator-context: three-lane access routing, infrastructure read tools in the map, and the two-layer context model (brief.md + profile.yaml).
• apollo-icp-builder, apollo-deliverability, apollo-go-live: wired to the new skills.

Full details in CHANGELOG.md.

v1.0.0 — Launch

The launch release. 13 skills turning the raw Apollo MCP into a guided outbound motion, built and tested live on a real Apollo account.

See CHANGELOG.md for the full skill list.