crmkit is an agent-first, headless CRM designed to allow AI agents to manage sales pipelines and contacts via standard HTTP requests or MCP.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install crmkit
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install crmkit using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is crmkit?
crmkit is an innovative, headless CRM designed from the ground up to be operated entirely by AI agents. Without a traditional graphical interface, your autonomous agents and LLMs act as the primary interface, driving operations via a plain-text HTTP API or the Model Context Protocol (MCP).
By leveraging structured endpoints and prioritizing grepable, single-line text responses (or standard JSON format), crmkit eliminates manual data entry overhead. Developers looking to implement robust, autonomous sales pipelines and lead nurturing workflows can leverage these Openclaw Skills to connect AI agents directly to core business data and CRM functionalities.
crmkit Use Cases
- Autonomously logging and managing sales deals, company directories, and contact pipelines within agentic workflows.
- Syncing and upserting contact and company records automatically during customer interactions using Openclaw Skills.
- Implementing context-rich customer tracking inside conversational AI clients like Claude or ChatGPT via the Model Context Protocol (MCP).
- Orchestrating pull-based task scheduling and follow-up activities using agent-triggered reminders.
How crmkit Works
- The AI agent authenticates via an email-based One-Time Password (OTP) sequence or automatically initiates a connection using the standardized MCP server flow.
- The agent executes targeted commands via raw HTTP requests or the single generic
requesttool exposed over MCP. - The backend processes the queries, handling upserts, searching records, or sorting workspace entities dynamically.
- Responses are sent back to the agent in highly readable, grepable plain text or JSON format depending on requested content headers.
- The agent acts on the fetched data, tracking upcoming follow-ups, resolving gated escalation flows with user confirmation, or logging activity histories.
crmkit Setup
Integrating crmkit as part of your Openclaw Skills catalog is straightforward. Follow this secure authentication sequence to generate an access token for your agent:
1. Request a 6-Digit Email Verification Code
Send a POST request with the user's email:
curl -s -X POST https://api.crmkit.ai/auth/request \\
-H "Content-Type: application/json" \\
-d '{"email":"[email protected]"}'
2. Verify the Code and Mint a Bearer Token
Prompt the user for the received code, then submit it along with an optional session descriptive name:
curl -s -X POST https://api.crmkit.ai/auth/verify \\
-H "Content-Type: application/json" \\
-d '{"email":"[email protected]","code":"123456","token_name":"AI-Agent-Session"}'
Save the returned bearer token securely. Your agent must pass this token as an Authorization: Bearer <token> header on all subsequent API requests. Active sessions can be queried via GET /tokens and revoked using DELETE /tokens/{id}.
crmkit Data Schema & Taxonomy
Core Data Entities
crmkit utilizes a simplified, lightweight data model optimized for easy parser-based queries and grep-friendly analysis:
| Entity | Key Fields | Description |
|---|---|---|
| Contact | id, name, email, company_id, stage, tags, custom, follow_up_at |
Client details. Posts with existing emails trigger an automatic update (upsert) instead of duplicate creation. |
| Company | id, name, domain, custom |
Associated organizations. Supports automatic upserts by domain name. |
| Deal | id, title, amount_cents, currency, stage, contact_id, company_id |
Sales opportunities. Monetary figures are recorded in integer cents (e.g., 100000 is $1,000 USD). |
| Activity | id, kind, body, created_at |
Operational history entries (kinds: call, email, meeting, note, task). |
Dynamic Custom Fields
Any payload keys sent inside the custom object wrapper are preserved and returned exactly as-is. This enables completely dynamic, schema-less extensions without database migrations.
Query Syntax
Search and filter collections dynamically using URL parameters formatted as field=[operator:]value. Multiple queries are logically ANDed:
?stage=lead(defaults toeqcheck)?amount_cents=gte:100000(evaluates relative inequalities using operators likegt,gte,lt,lte,ne)?stage=in:lead,qualified(matches any value in a comma-delimited list)?follow_up_at=is:null(validates empty and non-empty dates viais:nullandnot:null)?search=acme(performs fuzzy keyword searches across key records)
crmkit Advanced Features
- Model Context Protocol (MCP) Host: Supports native Claude and ChatGPT integration using standard dynamic client registration and OAuth flows directly at
https://api.crmkit.ai/mcp. - Multi-Tenant Workspaces: Execute isolated workspace operations under a single account. Easily switch scopes using dedicated workspace tokens.
- Two-Step Escapes & Gated Confirmations: Restrict sensitive operations like deletions or role promotions behind an explicit confirmation step, returning an
escalation_requiredstatus until confirmed via emailed OTP. - Pull-Based Reminders: Track actions through automated pulling of past-due reminders by leveraging targeted
follow_up_atflags and the specialized/reminderscheck-in endpoint.
SKILL.md
Loading
Related Openclaw Skills
Psychology Statistical Analysis Assistant
A comprehensive statistical analysis tool optimized for social cognition, experimental psychology, and quantitative behavioral research.
auth-md-signup
An authorization and authentication skill that enables AI agents to securely complete consent-preserving, user-claimed signups via the auth.md protocol.
Psychology Literature Search Assistant
A zero-dependency Python tool to search psychology literature across free academic databases and export citations in APA, GB/T 7714, MLA, and Chicago styles.
Research Method Decision Assistant (StatTestAdvisor)
A Python-based decision-support system that automatically recommends optimal statistical tests, sample sizes, psychometrics, and mediation/moderation workflows.
Cozy Game Planner Skill
A structured planning skill that guides users with clear templates, checklists, and safe defaults tailored for cozy game design and management.
Unitree Hermes Colab
A specialized safety-gated Google Colab workflow that leverages Hermes Agent to analyze Unitree Robotics repositories, generate simulation runbooks, and perform safe log triage.
Featured*
