{"id":8427,"date":"2026-08-06T05:00:40","date_gmt":"2026-08-06T05:00:40","guid":{"rendered":"https:\/\/www.coffee.ai\/articles\/automate-crm-data-with-claude"},"modified":"2026-08-06T05:00:40","modified_gmt":"2026-08-06T05:00:40","slug":"automate-crm-data-with-claude","status":"publish","type":"post","link":"https:\/\/www.coffee.ai\/articles\/automate-crm-data-with-claude","title":{"rendered":"How to Automate CRM Data with Claude: 7 Steps"},"content":{"rendered":"<p><em>Written by: Doug Camplejohn, CEO &amp; Co-Founder, Coffee<\/em><\/p>\n<h2 id=\"key-takeaways\">Key Takeaways for RevOps Leaders<\/h2>\n<ul>\n<li>Manual CRM data entry costs sellers roughly two hours per week and creates incomplete records, stale pipelines, and unreliable forecasts.<\/li>\n<li>Claude can automate enrichment, scoring, meeting summaries, and pipeline reporting once connected to HubSpot or Salesforce via MCP servers.<\/li>\n<li>Production-grade workflows rely on strict JSON schemas, robust error handling with retries, least-privilege permissions, and PII masking before any LLM call.<\/li>\n<li>Custom Claude scripts and Zapier workflows add ongoing maintenance work around schema changes, prompt drift, and write-back reliability that grows with team size.<\/li>\n<li>Coffee removes these maintenance issues with an always-on CRM agent that handles enrichment, summarization, and write-back automatically. <a href=\"https:\/\/www.coffee.ai\/pricing\" target=\"_blank\"><strong>Connect Coffee once and let the agent manage CRM automation for you.<\/strong><\/a><\/li>\n<\/ul>\n<h2>Prerequisites Before You Connect Claude to Your CRM<\/h2>\n<p>Confirm these prerequisites before you start building any Claude CRM workflow:<\/p>\n<ul>\n<li>Google Workspace or Microsoft 365 access for email and calendar data ingestion<\/li>\n<li>A Claude API key or an active Claude Code environment via Anthropic&#8217;s developer console<\/li>\n<li>CRM admin rights in Salesforce (Enterprise Edition or above) or HubSpot<\/li>\n<li>Basic familiarity with JSON schemas and OAuth 2.0 flows<\/li>\n<li>A dedicated integration user or scoped service account, never a system administrator account<\/li>\n<\/ul>\n<p><a href=\"https:\/\/www.coffee.ai\/pricing\" target=\"_blank\"><strong>Coffee eliminates the need to manage these prerequisites manually, connecting to your CRM in one step and handling authentication, scoping, and schema alignment for you.<\/strong><\/a><\/p>\n<h2>Step 1: Connect Claude to HubSpot<\/h2>\n<p>HubSpot&#8217;s remote MCP server became generally available on April 13, 2026. This release created the fastest path to a Claude CRM data entry workflow. Three integration routes exist, each with different trade-offs:<\/p>\n<ul>\n<li><strong>Native connector:<\/strong> Available directly in Claude.ai settings. <a href=\"https:\/\/syncgtm.com\/blog\/claude-code-crm-integration\" target=\"_blank\" rel=\"noindex nofollow\">Setup takes under 10 minutes for non-technical users<\/a> and grants read and write access to contacts, companies, deals, tickets, quotes, invoices, and activities.<\/li>\n<li><strong>Remote MCP server:<\/strong> Create an MCP Auth App under Development \u2192 MCP Auth Apps, create a managed OAuth 2.1 connection to <code>https:\/\/mcp.hubspot.com<\/code> using the app&#8217;s client ID and secret with PKCE, then connect from any MCP-compatible client. All actions respect the authenticated user&#8217;s existing HubSpot permissions.<\/li>\n<li><strong>Zapier middleware:<\/strong> Suitable for teams without developer resources. Zapier triggers on CRM events such as new deals or stage changes and passes structured payloads to Claude via the Anthropic action. This route introduces higher latency and per-task costs.<\/li>\n<\/ul>\n<p>A minimal JSON schema for a HubSpot contact write-back should define required fields explicitly and set <code>additionalProperties<\/code> to <code>false<\/code>. This setting prevents unexpected keys from reaching the CRM. <a href=\"https:\/\/dancroak.com\/ai\/claude\" target=\"_blank\" rel=\"noindex nofollow\">Requiring a per-job schema constant on every Claude API call keeps output consistent across batch runs.<\/a><\/p>\n<p><strong>Note:<\/strong> If a HubSpot account has sensitive data enabled, Activity objects are blocked from MCP server access. Plan field-level permissions before go-live so workflows do not fail silently.<\/p>\n<h2>Step 2: Connect Claude to Salesforce<\/h2>\n<p>Salesforce Hosted MCP Servers became generally available on April 29, 2026 in every Enterprise Edition org and above. Use this sequence to connect Claude:<\/p>\n<ol>\n<li>Navigate to Setup \u2192 API Catalog \u2192 MCP Servers and enable the server for your org.<\/li>\n<li>Create an External Client App with the <code>mcp_api<\/code> and <code>refresh_token<\/code> OAuth scopes.<\/li>\n<li>Configure your MCP client, such as Claude Code or Claude.ai, using OAuth 2.0 with PKCE.<\/li>\n<li>Scope the connected app to the minimum objects required, such as accounts, contacts, and opportunities, using profiles, permission sets, and field-level security instead of a system administrator account.<\/li>\n<\/ol>\n<p>Every transaction runs under the authenticated user&#8217;s existing permissions, CRUD and FLS rules, and sharing model. This behavior ensures full audit trails without anonymous service accounts. The <code>@salesforce\/mcp<\/code> package exposes <a href=\"https:\/\/syncgtm.com\/blog\/claude-code-crm-integration\" target=\"_blank\" rel=\"noindex nofollow\">more than 60 tools across data operations, metadata, users, orgs, testing, code analysis, and DevOps Center<\/a>. This toolset gives RevOps teams a powerful foundation for complex SOQL-based workflows.<\/p>\n<h2>Step 3: Automate Lead Enrichment with Claude<\/h2>\n<p>Lead enrichment becomes the first high-return automation once CRM connectivity is live. Companies that implement AI-driven lead workflows report conversion rate improvements ranging from roughly 19\u201338 percent in median cases, with some reaching 2x or higher compared to traditional methods.<\/p>\n<p>A production-grade enrichment prompt for Claude Sonnet should follow this structure:<\/p>\n<ul>\n<li>System prompt: instruct Claude to act strictly as a JSON API, responding only with valid JSON or the fallback object <code>{\"error\": \"reason\"}<\/code> when it cannot fulfill a request.<\/li>\n<li>User prompt: pass the raw contact record as structured JSON, list the fields to enrich such as job title, company size, funding stage, and ICP fit score, and include the instruction: <em>&#8220;Use only the provided data. If a field cannot be determined, return null. Never omit required fields.&#8221;<\/em><\/li>\n<li>Response schema: define <code>required<\/code> fields, set <code>additionalProperties: false<\/code>, and use <code>null<\/code> for unknown values instead of omitting keys.<\/li>\n<\/ul>\n<p><a href=\"https:\/\/dancroak.com\/ai\/claude\" target=\"_blank\" rel=\"noindex nofollow\">Match model selection to task complexity: use claude-haiku for short scoring and classification, claude-sonnet as the default for nuanced extraction, and claude-opus for deep research tasks.<\/a><\/p>\n<p><a href=\"https:\/\/claw.ist\/structured-output-claude-json-mode\" target=\"_blank\" rel=\"noindex nofollow\">Implement a retry loop with a default maximum of three attempts and exponential backoff for rate-limit errors such as HTTP 429 and schema validation failures<\/a>. Log parsing issues so you can refine prompts over time. For missing OAuth scopes, confirm the connected app includes read access to the specific CRM objects before you retry.<\/p>\n<h2>Step 4: Implement AI-Driven Lead Scoring<\/h2>\n<p>Once enrichment runs reliably, layer scoring on top of the enriched data. Use Claude to translate firmographic and behavioral signals into a consistent numeric or letter grade. Keep the scoring prompt simple and explicit so RevOps can adjust thresholds without rewriting code.<\/p>\n<p>Define clear inputs such as enrichment fields, recent activity counts, and opportunity history. Then define outputs such as score, reason codes, and recommended next actions. This structure keeps scoring transparent for sales leaders and easier to tune over time.<\/p>\n<h2>Step 5: Ingest Meeting Transcripts and Write Summaries Back to Your CRM<\/h2>\n<p>Meeting transcripts contain the richest unstructured CRM data and often receive the least attention. A reliable transcript-to-summary prompt includes these elements:<\/p>\n<ul>\n<li>Meeting type, attendees, and their roles<\/li>\n<li>The instruction: <em>&#8220;The following is a transcript including verbal filler words and speaker labels. Ignore filler words such as um, like, and you know, and organize substantive content.&#8221;<\/em><\/li>\n<li>Explicit output sections: a three to five sentence summary, decisions made, action items formatted as Owner \u2014 Task \u2014 Deadline, and open questions<\/li>\n<li>The fallback rule: <em>&#8220;If no deadline is mentioned, mark it TBD.&#8221;<\/em><\/li>\n<\/ul>\n<p><a href=\"https:\/\/claude-me.com\/en\/practice\/claude-meeting-notes-to-action-items\" target=\"_blank\" rel=\"noindex nofollow\">Explicit formatting instructions like &#8220;Owner \u2014 Task \u2014 Deadline&#8221; make Claude&#8217;s output directly usable for CRM or Slack follow-up without extra reformatting.<\/a><\/p>\n<figure style=\"text-align: center\"><a href=\"https:\/\/www.coffee.ai\/pricing\" target=\"_blank\"><img decoding=\"async\" src=\"https:\/\/cdn.aigrowthmarketer.co\/1763678549697-4e8d65abe17d.gif\" alt=\"GIF of Coffee platform where user is using AI to prep for a meeting with Coffee AI\" style=\"max-height: 500px\" loading=\"lazy\"><\/a><figcaption><em>Automated meeting prep with Coffee AI CRM Agent<\/em><\/figcaption><\/figure>\n<p>For write-back, pass the structured JSON output to the CRM&#8217;s notes or activity endpoint through the MCP server or REST API. <a href=\"https:\/\/www.coffee.ai\/changelog\" target=\"_blank\">Coffee&#8217;s improved summary templates, released in November 2025, are customizable to match workflows and writable back to Coffee, HubSpot, or Salesforce<\/a>. This capability removes the need to build and maintain the transcript pipeline yourself.<\/p>\n<h2>Step 6: Generate Week-over-Week Pipeline Reports Automatically<\/h2>\n<p>Natural-language pipeline reporting shows Claude&#8217;s value clearly inside a Claude CRM data entry workflow. A compare-report prompt should follow this pattern:<\/p>\n<ul>\n<li>Pull the current pipeline snapshot and the prior week&#8217;s snapshot as structured JSON inputs.<\/li>\n<li>Instruct Claude: <em>&#8220;Compare these two pipeline snapshots. List deals that progressed, deals that stalled with no stage change in seven or more days, new additions, and deals at risk of slipping the quarter. Output as JSON with keys: progressed, stalled, new, at_risk.&#8221;<\/em><\/li>\n<li>Include a fallback: <em>&#8220;If a deal appears in only one snapshot, classify it as new or closed_lost based on stage.&#8221;<\/em><\/li>\n<\/ul>\n<p><a href=\"https:\/\/apollo.io\/insights\/how-does-ai-driven-lead-scoring-improve-conversion-rates\" target=\"_blank\" rel=\"noindex nofollow\">A May 2026 Gartner report found that AI saves sellers 4.8 hours per week on average across all use cases<\/a>, and <a href=\"https:\/\/syncgtm.com\/blog\/claude-code-crm-integration\" target=\"_blank\" rel=\"noindex nofollow\">teams that ship AI-connected CRM workflows report two to three times faster pipeline reporting cycles<\/a>. These gains compound when RevOps leaders receive consistent week-over-week views without manual exports.<\/p>\n<h2>Step 7: Add Error Handling, Permissions, and Compliance Controls<\/h2>\n<p>Governance sits alongside functionality as a core requirement before launch. <a href=\"https:\/\/vantagepoint.io\/blog\/sf\/anthropic\/deploy-claude-safely-salesforce-hubspot\" target=\"_blank\" rel=\"noindex nofollow\">Treat data scoping, access controls, masking of sensitive fields, and audit logging as non-negotiable safeguards that prevent data leakage and compliance violations.<\/a><\/p>\n<p>A production-grade compliance checklist covers several layers of control that work together as a complete framework:<\/p>\n<ul>\n<li><strong>Least-privilege access:<\/strong> Scope the integration user to only the CRM objects the workflow requires. <a href=\"https:\/\/vantagepoint.io\/blog\/ai\/connect-crm-to-claude\" target=\"_blank\" rel=\"noindex nofollow\">Use dedicated integration users or scoped private apps instead of organization-wide access<\/a> so any compromise has a limited blast radius.<\/li>\n<li><strong>Read-before-write:<\/strong> Start in read-only mode and confirm output quality. Then expand to write actions for low-risk fields first so you catch schema mismatches before they corrupt production data.<\/li>\n<li><strong>Human-in-the-loop:<\/strong> Require human review before any output reaches a customer or triggers a data change until the workflow proves reliable over multiple weeks.<\/li>\n<li><strong>Audit logging:<\/strong> <a href=\"https:\/\/claudereadiness.com\/blog\/mcp-security-considerations-enterprise\" target=\"_blank\" rel=\"noindex nofollow\">Capture authentication events, tool invocations with sanitized parameters, data access details, and administrative actions, and retain logs for one to three years based on compliance needs.<\/a> These logs provide traceability when questions arise.<\/li>\n<li><strong>PII masking:<\/strong> <a href=\"https:\/\/k3x.ai\/blog\/gdpr-compliant-ai-crm-setup-checklist\" target=\"_blank\" rel=\"noindex nofollow\">Mask personally identifiable information before sending it to any LLM, and audit prompt inputs regularly to verify masking effectiveness.<\/a><\/li>\n<li><strong>SOC 2 and GDPR alignment:<\/strong> Anthropic is SOC 2 Type II certified and does not train on API inputs or outputs by default. Confirm that your middleware and CRM vendors hold current certifications so the full stack meets your standards.<\/li>\n<li><strong>API key rotation:<\/strong> <a href=\"https:\/\/claudereadiness.com\/blog\/mcp-security-considerations-enterprise\" target=\"_blank\" rel=\"noindex nofollow\">Rotate keys at least every 90 days, or every 30 days for critical systems, using a secrets manager such as AWS Secrets Manager or HashiCorp Vault.<\/a><\/li>\n<\/ul>\n<h2>Claude Scripts vs. Coffee Agent: Choosing a Scalable Path<\/h2>\n<p>Custom Claude scripts deliver quick wins for single workflows, yet the maintenance burden grows as use cases multiply, CRM schemas change, and token costs rise. The table below compares one-off scripts, Zapier workflows, and the Coffee agent across four operational dimensions.<\/p>\n<table>\n<thead>\n<tr>\n<th>Dimension<\/th>\n<th>One-off Claude Script<\/th>\n<th>Zapier + Claude Workflow<\/th>\n<th>Coffee Agent<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td><strong>Maintenance burden<\/strong><\/td>\n<td>High, because schema changes and prompt drift require developer intervention. <a href=\"https:\/\/rainautomation.com\/blog\/custom-crm-build-process-what-happens-week-by-week\" target=\"_blank\" rel=\"noindex nofollow\">A typical Claude-CRM integration project requires four to eight weeks to build and deploy.<\/a><\/td>\n<td>Medium, because Zapier handles triggers but prompt logic and field mappings still require ongoing updates.<\/td>\n<td>Low, because Coffee manages schema alignment, prompt versioning, and CRM field mapping automatically.<\/td>\n<\/tr>\n<tr>\n<td><strong>Always-on execution<\/strong><\/td>\n<td>No, because scripts run on demand or via scheduled cron, which creates gaps in coverage when triggers fail.<\/td>\n<td>Partial, because workflows are event-driven but subject to Zapier task limits and latency.<\/td>\n<td>Yes, because <a href=\"https:\/\/www.coffee.ai\/changelog\" target=\"_blank\">Coffee&#8217;s agent runs continuously, ingesting emails, calendars, and transcripts without human intervention.<\/a><\/td>\n<\/tr>\n<tr>\n<td><strong>Structured and unstructured data<\/strong><\/td>\n<td>Structured only by default, with unstructured data requiring custom parsing middleware.<\/td>\n<td>Structured only, with unstructured data such as transcripts and email threads requiring additional steps.<\/td>\n<td>Both, because Coffee unifies structured CRM fields and unstructured data such as emails and call transcripts into one coherent view backed by a built-in data warehouse.<\/td>\n<\/tr>\n<tr>\n<td><strong>Write-back reliability to Salesforce or HubSpot<\/strong><\/td>\n<td>Manual, because a developer must handle retries, field validation, and permission errors for each workflow.<\/td>\n<td>Partial, because Zapier writes back but lacks field-level validation and audit logging at the CRM layer.<\/td>\n<td>Production-grade, because <a href=\"https:\/\/www.coffee.ai\/changelog\" target=\"_blank\">Coffee writes enriched summaries and pipeline data back to HubSpot or Salesforce with deep integration knowledge of quotas, forecasting, and required fields.<\/a><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p><a href=\"https:\/\/www.coffee.ai\/pricing\" target=\"_blank\"><strong>Ready to eliminate the maintenance burden shown above? Coffee delivers production-grade reliability and always-on execution without custom code.<\/strong><\/a><\/p>\n<h2>Scaling Claude and Coffee for Small Teams and Mid-Market Orgs<\/h2>\n<p>The seven steps above apply across team sizes, yet the highest-impact starting point changes by context.<\/p>\n<p><strong>Small teams with 1\u201320 people:<\/strong> Automatic contact and company creation from Google Workspace or Microsoft 365 usually delivers the fastest win. Coffee&#8217;s agent scans emails and calendars to populate the CRM without manual input, replacing the spreadsheet and Notion shadow CRM that most early-stage teams rely on.<\/p>\n<figure style=\"text-align: center\"><a href=\"https:\/\/www.coffee.ai\/pricing\" target=\"_blank\"><img decoding=\"async\" src=\"https:\/\/cdn.aigrowthmarketer.co\/1763678186019-5cc1a76ac78e.gif\" alt=\"Build people lists automatically with Coffee AI CRM Agent\" style=\"max-height: 500px\" loading=\"lazy\"><\/a><figcaption><em>Build people lists automatically with Coffee AI CRM Agent<\/em><\/figcaption><\/figure>\n<p><strong>Mid-market teams on Salesforce or HubSpot:<\/strong> The Companion App model deploys Coffee as an intelligent layer on top of the existing system of record. A single authentication allows the agent to sync data, enrich records, and write insights back without migrating away from Salesforce or HubSpot. Coffee&#8217;s Visitor Identification feature adds another dimension: a single tracking pixel turns anonymous website traffic into named prospects with inferred name, title, email, and LinkedIn profile, with Suggested Leads that match a defined buyer persona surfaced in real time via Slack.<\/p>\n<p>For natural-language list building, Coffee&#8217;s List Builder accepts commands such as <em>&#8220;Find me VPs of Sales in North America at companies with $10M+ funding using Salesforce&#8221;<\/em> and executes the outbound workflow through integrated enrichment. This flow replaces the manual ZoomInfo export and Apollo sequence that most mid-market teams currently stitch together.<\/p>\n<figure style=\"text-align: center\"><a href=\"https:\/\/www.coffee.ai\/pricing\" target=\"_blank\"><img decoding=\"async\" src=\"https:\/\/cdn.aigrowthmarketer.co\/1763678641499-bad085f8165f.gif\" alt=\"Building a company list with Coffee AI\" style=\"max-height: 500px\" loading=\"lazy\"><\/a><figcaption><em>Building a company list with Coffee AI<\/em><\/figcaption><\/figure>\n<h2>Validation: How to Measure CRM Automation Success<\/h2>\n<p>Three measurement categories confirm that the automation delivers meaningful value:<\/p>\n<ul>\n<li><strong>Data-quality audits:<\/strong> Run monthly checks on field completion rates for contacts, companies, and opportunities. CRM data decays at 30 percent per year, so a rising completion rate after automation confirms that the agent fills gaps humans previously left.<\/li>\n<li><strong>Time recovered per rep:<\/strong> The target benchmark for CRM automation specifically is 8\u201312 hours per rep per week returned to selling activity. A Gartner study reported in 2024 found that sellers that partner with AI are 3.7 times more likely to hit quota compared to those who do not.<\/li>\n<li><strong>Pipeline accuracy:<\/strong> Compare forecast-to-close variance before and after automation. Coffee&#8217;s Pipeline Compare feature visualizes week-over-week changes such as progressed deals, stalled opportunities, and new additions, turning pipeline reviews from interrogation sessions into strategic discussions without manual CSV exports.<\/li>\n<\/ul>\n<h2>Frequently Asked Questions<\/h2>\n<h3>How long does initial setup take?<\/h3>\n<p>For teams using Coffee as a Companion App on HubSpot or Salesforce, setup requires a single authentication step. The Coffee agent begins ingesting emails, calendars, and meeting data immediately after connection, with no developer work required. Teams building custom Claude scripts from scratch should budget four to eight weeks for use-case definition, middleware development, testing, and deployment. Coffee removes that build time entirely.<\/p>\n<h3>Is my CRM data secure when using Claude?<\/h3>\n<p>Anthropic does not train on API inputs or outputs by default. Anthropic holds SOC 2 Type II certification and supports EU data residency for enterprise accounts. Coffee is also SOC 2 Type II and GDPR compliant, and customer data is never used to train public models. Best practice for any Claude CRM integration is to mask PII before it reaches the model, scope the integration user to the minimum necessary CRM objects, and maintain audit logs of every data access event.<\/p>\n<h3>When should I move from Zapier scripts to a full agent?<\/h3>\n<p>Zapier workflows work well for single, well-defined triggers with low data volume. The signal to move to a full agent appears when any of the following occur: more than two or three Zapier workflows touch the same CRM objects, prompt logic requires updating more than once per quarter because of schema or process changes, unstructured data such as transcripts or email threads needs processing alongside structured fields, or write-back errors appear frequently enough to require manual remediation. At that point, the maintenance cost of the script layer exceeds the cost of a production-grade agent.<\/p>\n<h3>Can Coffee work alongside my existing Salesforce or HubSpot instance?<\/h3>\n<p>Yes. Coffee&#8217;s Companion App model is designed specifically for teams committed to Salesforce or HubSpot. A single authentication deploys the Coffee agent as an intelligent layer on top of the existing system of record. The agent handles data enrichment, meeting summarization, activity logging, and pipeline reporting, then writes structured outputs back to the primary CRM. No migration is required, and Coffee has deep integration knowledge of Salesforce quotas, forecasting requirements, and required fields, areas where newer AI-native CRM alternatives frequently fall short.<\/p>\n<h2>Conclusion: Move from Scripts to Reliable CRM Automation<\/h2>\n<p>The seven steps in this guide, from connecting Claude to HubSpot and Salesforce via MCP through enrichment, scoring, transcripts, pipeline reporting, and compliance, form a complete path to a working Claude CRM data entry workflow. Each step delivers standalone value. Together, they eliminate the 8\u201312 hours per rep per week mentioned earlier and create the clean, structured data that accurate forecasting requires.<\/p>\n<p>Custom scripts and Zapier workflows provide a useful starting point but become a liability at scale. Schema drift, prompt maintenance, missing unstructured data support, and unreliable write-back combine into a second data-quality problem layered on top of the first. Coffee&#8217;s always-on agent solves the underlying issue of good data in and good data out without requiring your team to build or maintain the integration layer.<\/p>\n<p><a href=\"https:\/\/www.coffee.ai\/pricing\" target=\"_blank\"><strong>Start automating your CRM data at production scale with Coffee today.<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Learn how to automate CRM data with Claude using MCP, lead scoring &amp; enrichment. Coffee&#8217;s always-on CRM agent handles it all\u2014no maintenance required.<\/p>\n","protected":false},"author":11,"featured_media":8426,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"inline_featured_image":false,"footnotes":""},"categories":[1],"tags":[],"class_list":["post-8427","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/www.coffee.ai\/articles\/wp-json\/wp\/v2\/posts\/8427","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.coffee.ai\/articles\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.coffee.ai\/articles\/wp-json\/wp\/v2\/types\/post"}],"replies":[{"embeddable":true,"href":"https:\/\/www.coffee.ai\/articles\/wp-json\/wp\/v2\/comments?post=8427"}],"version-history":[{"count":0,"href":"https:\/\/www.coffee.ai\/articles\/wp-json\/wp\/v2\/posts\/8427\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.coffee.ai\/articles\/wp-json\/wp\/v2\/media\/8426"}],"wp:attachment":[{"href":"https:\/\/www.coffee.ai\/articles\/wp-json\/wp\/v2\/media?parent=8427"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.coffee.ai\/articles\/wp-json\/wp\/v2\/categories?post=8427"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.coffee.ai\/articles\/wp-json\/wp\/v2\/tags?post=8427"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}