Zevari Workflow Guide

Copy Prompts

Paste these prompts into Claude Code, Codex, Cursor, VS Code, Conductor, or another coding agent. Start with the business workflow prompt when you want the agent to audit your goals and recommend the right Zevari workflows before implementation.

• Ask Claude Code or Codex to design your Zevari workflows

  Use Zevari docs and help me design the right GTM workflows for my business before writing code. Business context: [describe what we sell, who we sell to, current GTM motion, team size, CRM/data tools, and constraints] Goals: [pipeline, lead generation, ICP scoring, campaign creation, follow-up, content, CRM sync, or other goals] Current data sources: [Apify datasets, CRM, Apollo, Clay, Google Sheets, CSV files, local files, or unknown] Preferred operating mode: [manual in Claude/Cursor/VS Code/Conductor/Codex, scheduled automation, or hybrid] Requirements: 1. Read https://docs.zevari.ai/help/workflows and https://docs.zevari.ai/mcp/reference before recommending tools, workflows, or automation. 2. Audit my business goals, ICP, current lead sources, CRM/data structure, sales process, content motion, and follow-up process. 3. Use Zevari through MCP/API only when I have connected authentication and the documented workflow supports that action. 4. Identify the highest-value Zevari workflows for my situation, then rank them by impact, complexity, data readiness, and risk. 5. For each recommended workflow, define inputs, outputs, source of truth, write-back destination, approval points, failure handling, and whether it should be manual, scheduled, or hybrid. 6. Ask for missing business details only when needed to avoid guessing. 7. Do not invent Zevari tools, endpoints, approval behavior, schedules, hidden states, or capabilities. 8. Stop before any send, campaign activation, delete, export, or bulk change unless the documented Zevari approval path is explicit. 9. Produce a practical implementation plan for Claude Code, Codex, Cursor, VS Code, or Conductor to install the workflows safely.

• Ask an AI IDE to install a GitHub Actions scheduler

  Use Zevari docs and help me install a scheduled GitHub Actions workflow for this Zevari job. Goal: [describe the Zevari workflow to run] Source data: [Apify dataset, CRM segment, CSV path, Apollo/Clay export, or other source] Destination: [Zevari list, CRM fields, Slack channel, output CSV, or review queue] Schedule: [daily, weekdays, hourly, weekly, or specific UTC cron] Requirements: 1. Read https://docs.zevari.ai/help/workflows and https://docs.zevari.ai/mcp/reference before writing code. 2. Create a script that is idempotent and safe to re-run. 3. Store Zevari API keys and external app tokens as GitHub Actions secrets, not in code or prompts. 4. Add both a scheduled trigger and workflow_dispatch for manual runs. Avoid minute 0 because GitHub scheduled workflows can be delayed or dropped during high-load times such as the start of the hour. 5. Log the run summary without printing secrets or private lead data. 6. Add a least-privilege permissions block, usually permissions: contents: read for a read-only scheduled runner. 7. Stop before any Zevari action that requires human approval, unless the documented API/MCP flow explicitly supports the approval path. 8. Show me the files you changed and how to test the workflow locally before pushing.

• Ask an AI IDE to install a local cron job

  Use Zevari docs and help me install a local scheduled job for this Zevari workflow. Goal: [describe the workflow] Local project path: [path] Data source: [CSV path, local JSON file, Apify API, CRM API, or other source] Schedule: [cron expression or plain English cadence] Requirements: 1. Read https://docs.zevari.ai/help/workflows and https://docs.zevari.ai/mcp/reference first. 2. Write a small script I can run manually before adding cron. 3. Load secrets from environment variables or a local secret manager, never from the script body. 4. Write outputs to a separate file or destination and include last_processed_at or an idempotency key. 5. Add logging so I can tell what happened after each run. 6. Give me the exact crontab, launchd, or systemd timer command for my OS. 7. Do not execute sends, deletes, campaign activation, or bulk changes without the documented Zevari approval path.

• Ask an AI assistant to design a no-code scheduler

  Use Zevari docs and design a no-code scheduler for this workflow. Goal: [describe the workflow] Scheduler: [n8n, Make, Zapier, CRM workflow, Apify schedule, or not sure] Source: [where records come from] Destination: [where results should be written] Requirements: 1. Read https://docs.zevari.ai/help/workflows and https://docs.zevari.ai/mcp/reference. 2. Break the workflow into trigger, read, transform, Zevari call, review gate, write-back, and notification steps. 3. Identify which steps can run automatically and which must stop for human review. 4. Define the fields each step reads and writes. 5. Include retry, dedupe, and failure-notification rules. 6. Do not invent Zevari tools, endpoints, schedules, approval behavior, or hidden workflow states.

How Workflows Run

Zevari workflows can run manually in an AI IDE or automatically through a scheduler. Start new workflows manually in Cursor, VS Code, Conductor, Claude, ChatGPT, or another MCP-capable client. Once the process is stable, schedule the repeatable read, score, draft, sync, and reporting steps through GitHub Actions, local cron, n8n, Make, Zapier, CRM automation, Apify schedules, or a server runner.

• Manual in an AI IDE

  Run the workflow interactively from Cursor, VS Code, Conductor, Claude, ChatGPT, or another MCP-capable client. This is best when the operator wants to inspect research, adjust copy, choose targets, or approve sensitive actions in real time.

• Scheduled with a runner

  Run the same workflow on a schedule through GitHub Actions, local cron, n8n, Make, Zapier, a server job, or another scheduler. The scheduled job reads the source data, calls Zevari or asks an agent to call Zevari, writes outputs back to the chosen system, and stops at required approval boundaries.

• Hybrid control

  Let the scheduler handle low-risk reads, scoring, list cleanup, reporting, and draft creation, then route sends, campaign activation, deletes, exports, and bulk changes back to a human review step.

Operating Rules

• Load docs first

  Every AI-run workflow should read https://docs.zevari.ai/mcp/reference before it names Zevari tools, fields, approvals, schedules, or workflow states.

• Use a real data source

  The source of truth can be Apify, a CRM, Apollo, Clay, Airtable, Google Sheets, a CSV on a local drive, or another data store. Define stable IDs, ownership, field mapping, and write-back rules before automation.

• Keep secrets out of code

  Use GitHub Actions secrets, environment variables, secret managers, or encrypted automation-platform secrets. Do not store API keys in prompts, CSV files, scripts, or logs.

• Make jobs idempotent

  Use stable IDs, processed timestamps, status fields, and dedupe keys so reruns do not duplicate leads, campaign targets, CRM notes, or sends.

• Stop at approval boundaries

  Automate research, scoring, drafts, reports, and review queues. Sensitive sends, campaign activation, deletes, exports, and bulk changes must follow documented Zevari approval behavior.

Setup and Inputs

These workflows establish the operating context, source data, and target records that later research, scoring, outreach, and automation depend on.

• Workspace and Context Setup

  Outcome: A confirmed Zevari workspace, organization, LinkedIn sender, ICP, voice, Library assets, templates, and safety posture. Inputs: Workspace or organization, LinkedIn sender, ICP and offer, Voice DNA or writing examples, Templates and CRM field map. Manual run: Ask the AI assistant to read the Zevari MCP reference, confirm active context, load Library assets, and summarize what it can safely do before any write. Scheduled run: Schedule a daily context check that verifies the intended workspace and sender, refreshes known Library/ICP state, and reports setup gaps before running downstream jobs. Stack: Cursor, VS Code, Conductor, Claude, ChatGPT, Zevari app, CRM, Notion, Slack.

• Lead Source and Input

  Outcome: A normalized intake path for LinkedIn URLs, Apollo lists, Clay tables, Apify datasets, CRM segments, CSV files, or manual prospects. Inputs: Source system, Lead identifier, Required fields, Owner or workspace, Destination list or pipeline stage. Manual run: Paste URLs, CSV rows, CRM records, or list names into the AI session and ask it to normalize the source before research begins. Scheduled run: Run a scheduled importer that reads the source of truth, deduplicates by LinkedIn URL or provider identity, and saves new or changed leads into Zevari. Stack: Apify, Apollo, Clay, HubSpot, Salesforce, Pipedrive, Google Sheets, CSV, local files, Zapier, Make, n8n.

• Lead List Cleanup and Deduplication

  Outcome: Clean target lists with duplicate profiles merged, invalid records removed, and suppressed contacts excluded. Inputs: Raw list, Suppression rules, LinkedIn URL or provider ID, Email when available, Owner workspace. Manual run: Ask the AI to inspect a list, identify duplicates and missing fields, then propose cleanup actions for review. Scheduled run: Run cleanup before every scheduled research or campaign job so stale, duplicate, opted-out, or incomplete records do not enter execution. Stack: Zevari target lists, CRM lists, Apify datasets, CSV, Apollo exports, Clay tables.

Prospecting and Qualification

These workflows turn markets, lists, posts, and signals into researched and ranked prospects.

• LinkedIn Prospect Discovery

  Outcome: A raw prospect pool built from LinkedIn people, company, post, commenter, or keyword searches. Inputs: Persona, Industry, Geography, Seniority, Company size, Search keywords. Manual run: Use an AI session to explore search criteria, inspect examples, and tune the search before saving a list. Scheduled run: Schedule recurring discovery for defined personas or keywords, then write new candidates into an intake list for review. Stack: Zevari LinkedIn search, Sales Navigator, Apify actors, Apollo, Clay, CRM segments.

• Prospect Research and Qualification

  Outcome: Research summaries with role, company, activity, relevance, and recommended next action. Inputs: Target list, Research depth, ICP rule, Offer or CTA, Exclusion rules. Manual run: Ask the AI to research a named account or short list, then review the reasoning before saving qualification state. Scheduled run: Run nightly research against new intake leads and store summaries, missing-data flags, and qualification decisions for the next operator review. Stack: Zevari, LinkedIn profiles, company pages, CRM records, Apollo, Clay, public web research.

• ICP Scoring

  Outcome: A Zevari 1-5 ICP score, fit rationale, and include or exclude decision for each lead. Inputs: Target industry, Target role, Product offering, Must-have criteria, Nice-to-have signals. Manual run: Have the AI score a reviewed list and explain exactly why each person is a 1, 2, 3, 4, or 5 fit. Scheduled run: Score every new intake lead daily, then push high-fit leads to a qualified list while routing low-confidence scores to review. Stack: Zevari ICP scoring, CRM properties, Apollo/Clay enrichment, CSV imports.

• Audience Gap Intelligence

  Outcome: A ranked gap audience from people who engage with competitors, partners, creators, or category posts but not with you. Inputs: Baseline post URLs, Comparison post URLs, Audience tags, ICP scoring rule. Manual run: Paste post URLs and ask Zevari to compare commenters, explain overlap, and rank the missing audience. Scheduled run: Schedule periodic scans of selected posts or Apify datasets, compare against your baseline audience, and create a weekly gap report. Stack: Zevari, LinkedIn post commenters, Apify, Notion, Slack, CRM target lists.

• Signal Detection and Scoring

  Outcome: Scored buying or engagement signals that tell the operator whether to research, draft outreach, create content, or stand down. Inputs: Signal keywords, Companies or leads, Post sources, Minimum score, Action policy. Manual run: Review a batch of signals with the AI and choose which ones should move to research or outreach. Scheduled run: Run scheduled signal scans and route only signals scoring 3+ into research, outreach drafting, CRM tasks, or Slack alerts. Stack: Zevari signals, Apify, LinkedIn posts, CRM watchlists, Slack, n8n, Make.

• Mutual Intro Finder

  Outcome: A ranked list of mutual connections and safe opener lines that reference only verified mutuals. Inputs: Target LinkedIn URL, Known close relationships, Evidence mode, Outreach goal. Manual run: Ask the AI to inspect one target and rank intro paths before deciding whether to ask for an intro. Scheduled run: Run on high-value qualified accounts and save warm-intro evidence for operator review before outreach. Stack: Zevari mutual connections, inbox evidence, CRM account priority, Slack review.

Outreach and Campaigns

These workflows move qualified leads into reviewed messages, warm-up paths, campaigns, or multichannel execution.

• Research-to-Outreach

  Outcome: A research-backed communication plan plus LinkedIn, InMail, or email drafts. Inputs: Prospect or list, Reason for outreach, Offer or CTA, Preferred channel, Tone constraints. Manual run: Run this interactively for named prospects when message quality matters more than volume. Scheduled run: Draft messages for newly qualified leads each morning, then send the drafts to a human queue for approval or campaign enrollment. Stack: Zevari Library, templates, CRM notes, Apollo/Clay enrichment, email tools.

• Warm-Up-to-Connect

  Outcome: A safer LinkedIn path from research to profile views, engagement, and connection requests. Inputs: Qualified prospects, Warm-up duration, Connection-note angle, Minimum ICP score. Manual run: Ask the AI to propose warm-up steps and connection notes, then review before creating or activating anything. Scheduled run: Schedule warm-up planning for active prospects while keeping connection requests and comments behind approval when required. Stack: Zevari campaigns, LinkedIn activity data, CRM segments, Slack approvals.

• Campaign Creation

  Outcome: A verified draft or active campaign with targets, step config, generated copy, delays, and safety-aware pacing. Inputs: Campaign goal, Target list, Template or framework, Resource URL, Activation preference. Manual run: Use Cursor, VS Code, Conductor, Claude, or ChatGPT to create a draft, inspect the saved campaign, and approve activation. Scheduled run: Create draft campaigns from scheduled qualification outputs, then notify the operator to review and activate. Stack: Zevari campaigns, outreach templates, CRM lists, Apollo/Clay enriched fields, Slack or email notifications.

• Multi-Channel Outreach

  Outcome: Coordinated LinkedIn and email steps routed by lead activity, email availability, and channel fit. Inputs: LinkedIn target data, Email enrichment, Channel rules, CTA, Sequence policy. Manual run: Ask the AI to segment active, moderate, and inactive LinkedIn users and draft the appropriate channel sequence. Scheduled run: Run scheduled channel routing, use Apollo or Clay for email enrichment, create Zevari LinkedIn drafts, and hand email steps to Smartlead, Instantly, or another sender. Stack: Zevari, Apollo, Clay, Smartlead, Instantly, HubSpot, Salesforce, Zapier, Make, n8n.

• Template-Driven Messaging

  Outcome: Messages generated from reusable frameworks such as pain-led, value-led, authority-led, and Hook-Starter variants. Inputs: Template slug, Prospect context, Campaign goal, Voice constraints, CTA. Manual run: Browse templates, fork or customize one, generate samples, and review the messaging logic before using it. Scheduled run: Apply approved templates to qualified leads on a schedule, then store drafts and template attribution for later outcome analysis. Stack: Zevari Library, outreach templates, campaign targets, CRM fields.

• Campaign Activation and Safety Review

  Outcome: A final preflight that checks target identity, sender context, step content, approvals, working hours, and safety state. Inputs: Campaign ID, Activation preference, Approval policy, Sender account, Known risks. Manual run: Have the AI validate the campaign, summarize risks, and ask for approval before activation. Scheduled run: Run scheduled campaign preflight checks and send a daily list of campaigns that are ready, blocked, or unsafe to activate. Stack: Zevari Safety Center, confirmations, campaigns, Slack, email digest.

Follow-Up, Pipeline, and Content

These workflows keep replies, pipeline state, meetings, and content connected to the original GTM motion.

• Campaign Monitoring and Recovery

  Outcome: A clear view of stuck campaigns, blocked targets, missing content, safety pauses, and safe recovery actions. Inputs: Campaign IDs, Date range, Failure state, Recovery policy. Manual run: Ask the AI to inspect campaign health and propose recovery steps without bypassing safety or stale approvals. Scheduled run: Schedule a campaign health report that sends blockers, ready-to-resume campaigns, and required human decisions to Slack or email. Stack: Zevari campaign health, Safety Center, Slack, email, CRM tasks.

• Inbox Triage and Hot Lead Routing

  Outcome: Classified replies, hot-lead flags, ICP-gated routing, and recommended response paths. Inputs: Inbox window, Lead context, ICP gate, Routing rules, CRM stage map. Manual run: Open the reply queue with the AI, inspect the classifier reasoning, and choose the next action. Scheduled run: Run scheduled or webhook-triggered reply classification and route high-intent replies to Slack, CRM tasks, or meeting workflows. Stack: Zevari inbox, CRM, Slack, booking tools, webhook automation.

• Quiet Reply Re-Engagement

  Outcome: One context-aware follow-up for people who replied, went quiet, and are not already active in a campaign. Inputs: Quiet window, Lead or segment, Follow-up goal, Tone constraints. Manual run: Ask the AI to find one quiet conversation, read the local thread, draft one reply, and stage confirmation. Scheduled run: Run a scheduled scan for quiet replies and create a review queue of proposed one-off follow-ups. Stack: Zevari inbox, leads, confirmations, CRM tasks, Slack review.

• CRM Pipeline Sync

  Outcome: Qualified leads, activities, replies, meetings, opportunities, and campaign outcomes reflected in the CRM. Inputs: CRM object map, Lead IDs, Activity types, Stage rules, Attribution fields. Manual run: Ask the AI to prepare CRM updates from confirmed Zevari evidence and show the proposed changes before writing. Scheduled run: Sync daily or webhook-triggered changes from Zevari to HubSpot, Salesforce, Pipedrive, or another CRM. Stack: Zevari API, HubSpot, Salesforce, Pipedrive, Zapier, Make, n8n.

• Meeting and Booking Workflow

  Outcome: Positive replies moved toward a booking link, logged meeting, or CRM opportunity with attribution. Inputs: Positive reply, Booking link, Lead context, Campaign attribution, CRM stage. Manual run: Have the AI draft the right booking reply and log the meeting or next-step proposal after review. Scheduled run: Route positive replies into booking tasks, attribute calendar events back to leads and campaigns, and update CRM stages. Stack: Zevari, Calendly, SavvyCal, HubSpot, Salesforce, Slack.

• LinkedIn Content Drafting and Scheduling

  Outcome: Drafted posts saved to Zevari for review, media upload, scheduling, and eventual publishing. Inputs: Topic, Audience, Voice, Number of posts, Schedule shape. Manual run: Ask the AI to draft posts and save them as Zevari content rows, then upload media and schedule in the dashboard. Scheduled run: Generate weekly drafts on a schedule, save them for review, and notify the operator to attach media and choose publish times. Stack: Zevari content, Library, Notion, Google Docs, Slack, LinkedIn.

• Lead Magnet Creation and Promotion

  Outcome: Lead magnet ideas, assets, promotion posts, commenters, and follow-up lists connected to the GTM workflow. Inputs: Topic, ICP, Asset format, Promotion CTA, Comment keyword. Manual run: Use the AI to generate the asset idea, draft promotional posts, and plan how commenters enter follow-up. Scheduled run: Schedule recurring idea generation, post draft creation, commenter collection, and lead routing into Zevari or CRM lists. Stack: Zevari content, Apify commenters, Beehiiv, ConvertKit, Mailchimp, CRM, Zapier.

• Industry Pulse and Content Intelligence

  Outcome: A recurring market brief with trends, competitor movement, content opportunities, and outreach triggers. Inputs: Industry keywords, Competitor profiles, Watchlist, Report cadence, Output destination. Manual run: Ask the AI to research a market or topic and convert findings into posts, outreach angles, or campaign ideas. Scheduled run: Run weekly scans and publish the brief to Slack, Notion, email, or a CRM account research field. Stack: Zevari, LinkedIn posts, Apify, Notion, Slack, email digests.

Stack Orchestration

These workflows connect Zevari to the rest of the GTM stack so one AI-controlled process can read data, call tools, write outputs, and preserve human review.

• AI IDE Orchestration

  Outcome: A repeatable agent runbook that uses Cursor, VS Code, Conductor, Claude, or ChatGPT as the workflow control room. Inputs: Goal, Zevari docs, API key or MCP auth, Source systems, Approval rules. Manual run: Run the workflow in an AI IDE, inspect every tool call, and ask the assistant to explain the exact next action before writes. Scheduled run: Have the IDE or agent scaffold scripts, GitHub Actions workflows, local cron entries, or n8n flows that run the same process on schedule. Stack: Cursor, VS Code, Conductor, Claude, ChatGPT, Codex, GitHub Actions, cron.

• External Data Store Workflow

  Outcome: A clear source-of-truth pattern for using Apify, CRM records, CSV files, or another system as lead storage. Inputs: Data source, Primary key, Sync direction, Field map, Conflict policy. Manual run: Ask the AI to read the source, show the field map, and propose how records should flow into or out of Zevari. Scheduled run: Run scheduled sync jobs that read from the chosen store, call Zevari for research/scoring/drafts, and write results back with timestamps. Stack: Apify datasets, HubSpot, Salesforce, Airtable, Google Sheets, CSV, local filesystem, S3/R2.

Scheduler Options

A scheduled Zevari workflow usually has a scheduler, a runner, a data source, and a write-back destination.

• GitHub Actions schedule

  Use a repository workflow with an `on.schedule` cron expression and a manual `workflow_dispatch` trigger. Keep API keys in GitHub Actions secrets, run from the default branch, avoid the start of the hour when choosing cron minutes, set least-privilege token permissions, and treat scheduled runs as best-effort jobs that should be idempotent. Best for: Teams that want versioned automation, logs, pull requests, and a simple hosted runner.

• Local cron

  Run a script on a trusted machine with `crontab`, launchd, systemd timers, or another local scheduler. Store secrets outside the script and log each run to a file or monitoring system. Best for: Operators with a stable workstation, server, or internal machine that already runs scheduled jobs.

• n8n, Make, or Zapier

  Use a no-code or low-code scheduler to trigger reads from CRM, Apify, Google Sheets, or webhooks, then call Zevari API endpoints or an agent-controlled bridge. Best for: Non-engineering GTM teams that want visual workflows and app connectors.

• CRM or data-platform automation

  Let HubSpot, Salesforce, Clay, Apollo, or Apify trigger the workflow when a record enters a segment, changes stage, or appears in a dataset. Best for: Teams that already use a CRM or enrichment platform as the operational source of truth.

• Server job or queue worker

  Run scheduled jobs in your own backend, queue, or worker infrastructure. Add retries, idempotency keys, audit logs, and explicit review queues for sensitive actions. Best for: Engineering teams that need deeper control, compliance, or higher-volume internal automation.

Data Source Patterns

Zevari does not require one CRM pattern. Choose a source of truth that gives the runner clear, current, and permissioned data.

• Apify as the lead warehouse

  Use Apify datasets for scraped LinkedIn search results, post commenters, competitor audiences, or recurring data collection. Zevari can then research, score, draft, and route the records. Best practice: Keep stable dataset IDs, store LinkedIn profile URLs, and write Zevari output fields such as score, status, last_processed_at, and next_action.

• CRM as source of truth

  Use HubSpot, Salesforce, Pipedrive, or another CRM as the canonical place for companies, contacts, stages, owners, and attribution. Best practice: Map Zevari lead IDs and LinkedIn URLs back to CRM record IDs, and avoid letting scheduled jobs overwrite human-owned CRM fields without review.

• CSV or local files

  Use local CSV files for simple or early workflows. This works well for operators who want an AI IDE to read a list, call Zevari, and write an output CSV. Best practice: Use separate input and output files, never store API keys in the CSV, and keep a processed timestamp so repeated cron runs do not duplicate work.

• Apollo or Clay enrichment

  Use Apollo or Clay to supply emails, firmographics, technographics, and enrichment fields while Zevari handles LinkedIn context, ICP scoring, and outreach preparation. Best practice: Let enrichment run before channel routing. Only enroll leads in email steps when email data is verified and consent/compliance rules are satisfied.

• Hybrid working set

  Use one system for raw lead intake, Zevari for LinkedIn research and execution state, and a CRM for pipeline records. Best practice: Define one primary key per record, store sync timestamps, and make scheduled jobs safe to re-run without duplicating leads or actions.

GitHub Actions Scheduled Runner

GitHub Actions scheduled workflows use POSIX cron syntax, run on the default branch, and use UTC. Avoid minute 0 because high-load times include the start of every hour. Add workflow_dispatch for manual runs. Set least-privilege permissions. Store Zevari API keys and external app tokens in GitHub Actions secrets and pass them through the secrets context.

• Example workflow

  name: Scheduled Zevari Runner on: schedule: - cron: "17 13 * * 1-5" workflow_dispatch: permissions: contents: read jobs: run: runs-on: ubuntu-latest steps: - uses: actions/checkout@v6 - uses: actions/setup-node@v6 with: node-version: "22" - run: npm ci - run: node scripts/scheduled-zevari-runner.mjs env: ZEVARI_API_KEY: ${{ secrets.ZEVARI_API_KEY }} CRM_API_KEY: ${{ secrets.CRM_API_KEY }}

Docs Links

• Help Center

  Human-facing guide to Zevari skills, workflows, agents, videos, safety, and support.

• Workflow Guide

  Manual and scheduled Zevari workflow guide for GTM operators and AI IDE orchestration.

• Prompting Guide

  Copy-ready prompts that require AI assistants to read the MCP reference before acting.

• Workspaces and Sender Seats

  Human-facing setup guide for workspaces, members, billing, add-on LinkedIn sender seats, and teammate LinkedIn connections.

• Safety Center

  Human-facing guide to Zevari safety guardrails, warm-up, pauses, blocked actions, and recovery state.

• Warm-Up

  Human-facing guide to LinkedIn sender warm-up and gradual activity ramping.

• MCP Reference

  Agent-facing tool schemas, capability contracts, examples, gotchas, and recovery guidance.

• API Reference

  Public REST API reference powered by the Zevari OpenAPI document.

• MCP Reference JSON

  Machine-readable MCP tool reference.

• LLMs Full

  AI-readable Zevari documentation bundle.

• Support

  Send bug reports, support handoffs, and feature requests.