Codex ZoomInfo Skills for GTM Teams

By Rome Thorndike | June 2026

What This Setup Replaces

ZoomInfo is the dominant enterprise contact data platform for B2B GTM. Most ZoomInfo users still operate it through the web UI: log in, search, export to CSV, paste into the CRM, run outbound. The manual loop costs hours per week per rep.

Codex skills that wrap ZoomInfo data turn the loop into an agent workflow. The agent pulls the data, scores it, decides what to do, writes to the CRM. The human reviews the queue rather than running the search.

For the broader Codex CLI patterns, see the Codex CLI guide.

Step 1: Wire ZoomInfo Access

The cleanest path is a custom MCP wrapper around ZoomInfo's REST API. The wrapper exposes three operations.

company_search. Takes industry, employee size, geography, technology, and intent filters. Returns matching companies with firmographics.

contact_search. Takes a company domain plus title filters. Returns matching contacts with verified emails and direct phones.

intent_pull. Takes a topic list. Returns accounts with intent activity in last 7 days, with intent score and source.

The wrapper authenticates with your ZoomInfo OAuth token, stored in ZOOMINFO_TOKEN environment variable. Configure in Codex config.toml as a stdio MCP server pointing at your wrapper script.

Step 2: Build the Account Research Skill

Trigger: "/zi-research [company]" or "look up [company] in ZoomInfo".

Steps. 1. Take the company name or domain. 2. Call ZoomInfo company_search to get the company record. 3. Call contact_search filtered to VP+ titles in revenue functions. 4. Score the account against ICP rules in AGENTS.md. 5. Output a brief with firmographics, top 3 buyer contacts, ICP score, and a recommended next step.

This is the entry-level skill. Most teams ship it first because it produces visible value (clean account briefs) without writing anything to the CRM.

Step 3: Build the Lead Scoring Skill

Trigger: scheduled daily run, or "/zi-score [account list]".

Steps. 1. Read a list of target accounts from the CRM. 2. For each account, call ZoomInfo company_search to refresh firmographics. 3. Score against the ICP rubric. 4. Write the score and tier to a custom CRM property. 5. Route Tier A accounts to a Slack channel for SDR attention.

This is the workflow that earns its keep at scale. A team enriching 500 accounts a week saves 8 to 12 hours of SDR time.

Step 4: Build the TAM Analysis Skill

Trigger: "/zi-tam [ICP definition]" or quarterly run.

Steps. 1. Take the ICP definition from AGENTS.md (or accept a custom one). 2. Call ZoomInfo company_search with the ICP filters and no result cap. 3. Cluster the results by sub-segment (industry, employee size, geography). 4. Output a TAM analysis with total account count, sub-segment breakdown, and the top 100 accounts by score.

Use this for territory planning, quarterly TAM refresh, and new-market evaluation. Less frequent than the other two skills but high value per run.

Step 5: Add the Intent Workflow

If your ZoomInfo subscription includes intent data, build the intent-prioritized prospecting workflow.

Steps. 1. Daily, call intent_pull for your tracked topics. 2. For each account showing intent, check ICP fit using the scoring skill. 3. For accounts with both intent and ICP fit, write to the prospecting queue with intent topic noted. 4. Sort the queue by combined score (intent intensity plus ICP fit).

SDRs work the queue starting with the top of the list. The pattern that works: filter intent-only accounts out of the SDR's view (they're noise without ICP fit) and surface intent-plus-fit prominently.

Cost and Operational Considerations

ZoomInfo API access is paid per query bucket. Most plans give 5,000 to 50,000 monthly credits depending on tier. The skills above consume credits per call: company_search costs 1, contact_search costs 2 to 5 depending on results, intent_pull costs more per account. Budget for 30 to 50% credit consumption on the skills if they're running at moderate volume.

Add credit usage logging to your wrapper so you can track per-skill consumption. If one skill is burning credits faster than expected, the wrapper logs surface it before your monthly quota runs out mid-month.

Related Patterns

For broader build patterns, see the Codex sales agent guide and Claude Code sales agent guide. For the alternative if you don't have ZoomInfo, see the Clay-based enrichment workflow. For underlying ZoomInfo capability, see the ZoomInfo review.

Authoritative References

For Codex MCP setup, see OpenAI's Codex MCP documentation. For ZoomInfo's API and Copilot, see the ZoomInfo Copilot page.

Frequently Asked Questions

Does ZoomInfo have an official MCP server for Codex?

ZoomInfo announced an MCP-compatible API surface in 2025 as part of the Copilot integration roadmap. The official server is in limited availability for enterprise customers as of mid-2026. Most teams use a custom wrapper around ZoomInfo's REST API instead, which exposes the same operations (contact search, company enrichment, intent data) through a Codex-compatible interface. The wrapper takes about 50 lines of Python using the Anthropic SDK template.

What ZoomInfo data is most useful for Codex workflows?

Three datasets. Contact data (verified emails, direct phones, titles) for outbound enrichment. Company firmographics (employee count, revenue, industry) for ICP scoring and TAM. Intent data (signals showing accounts researching your category) for prioritization. Most teams wire all three but use them in different workflows. Intent data drives the prioritization workflow. Contact data drives outbound enrichment. Firmographics drive account scoring.

How does Codex use ZoomInfo intent data?

The intent data API returns accounts showing recent research activity in topics you've defined as relevant. Codex reads this list daily, scores each account against your ICP rules, and writes the top-scoring accounts to a prospecting queue with the intent topic noted. Your SDRs work the queue starting with the accounts showing both ICP fit and intent. Intent signal alone is noisy; intent plus ICP fit is the workflow that converts.

Is ZoomInfo plus Codex cheaper than ZoomInfo plus their native Copilot?

Depends on volume and your existing tooling. ZoomInfo Copilot is bundled into the ZoomInfo subscription and handles common workflows without additional model spend. A Codex workflow on top of ZoomInfo adds model cost (typically $200 to $800/mo for moderate GTM use) but gives you flexibility on workflow design and integration with the rest of your stack. Most teams that already have a Codex or Claude Code workflow added find the marginal cost worth the consistency.

Should I use ZoomInfo or Clay for Codex-based account research?

Different strengths. ZoomInfo is stronger on enterprise contact data, direct phone numbers, and US-focused company data. Clay is stronger on flexibility, waterfall orchestration across many vendors, and SMB coverage. Most teams that have both use ZoomInfo for high-value enterprise account research where direct phones matter, and Clay for higher-volume SMB enrichment where flexibility matters. If you only have one, pick based on which segment dominates your ICP.

More Insights

Insights Index GTM Engineer Job Market: 2026 Analysis GTM Engineer Salary Trends: What the Data Shows GTM Tool Adoption: Which Tools Are Winning State of GTM Engineering 2026: Key Findings The Clay Ecosystem: Who Builds What The 2026 Outbound Stack: What Top GTM Engineers Use Clay Playbook: Advanced Workflows That Work LinkedIn Outreach for GTM Engineers: A Technical Guide Email Deliverability for GTM Engineers API Integration Patterns for GTM Engineers Data Enrichment Waterfall Strategy