Skip to main content

​
Quickstart: Your First Agent Request

This guide walks you through the basics of Agent-CoreX in under 5 minutes.

​
Step 1: Sign Up

Go to agent-corex.com and create a free account. You’ll get:
  • API key for authentication
  • Access to MCP marketplace (100+ servers)
  • 1,000 free API requests/month

​
Step 2: Install the SDK

npm install agent-corex
Or using Python: 
pip install agent-corex

​
Step 3: Make Your First Request

​
JavaScript/TypeScript

import { AgentCorex } from 'agent-corex';

const agent = new AgentCorex({
  apiKey: process.env.AGENT_COREX_API_KEY
});

// Step 1: Retrieve tools for a specific task
const tools = await agent.retrieveTools({
  query: "Create a GitHub PR and notify the team on Slack",
  topK: 5
});

console.log('Available tools:', tools);
// Output:
// [
//   { name: 'create-pull-request', score: 0.98 },
//   { name: 'list-repositories', score: 0.95 },
//   { name: 'send-slack-message', score: 0.92 },
//   { name: 'post-comment', score: 0.88 }
// ]

// Step 2: Execute a tool
const result = await agent.executeTool({
  toolName: 'create-pull-request',
  params: {
    repository: 'owner/repo',
    title: 'Auto-generated feature PR',
    body: 'This PR was created automatically by Agent-CoreX',
    fromBranch: 'feature/auto-deploy',
    toBranch: 'main'
  }
});

console.log('PR Created:', result);
// {
//   success: true,
//   pullRequestUrl: 'https://github.com/owner/repo/pull/123'
// }

// Step 3: Chain tools (bonus!)
await agent.executeTool({
  toolName: 'send-slack-message',
  params: {
    channel: '#deployments',
    message: `🚀 New PR created: ${result.pullRequestUrl}`
  }
});

​
Python

from agent_corex import AgentCorex

agent = AgentCorex(api_key="your_api_key")

# Retrieve tools
tools = agent.retrieve_tools(
    query="Create a GitHub PR and notify the team on Slack",
    top_k=5
)

print("Available tools:", tools)

# Execute a tool
result = agent.execute_tool(
    tool_name="create-pull-request",
    params={
        "repository": "owner/repo",
        "title": "Auto-generated feature PR",
        "body": "Created by Agent-CoreX",
        "from_branch": "feature/auto-deploy",
        "to_branch": "main"
    }
)

print("PR Created:", result)

​
cURL

# 1. Retrieve tools
curl -X GET "https://api.agent-corex.com/retrieve_tools?query=Create%20a%20GitHub%20PR" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json"

# 2. Execute a tool
curl -X POST "https://api.agent-corex.com/execute_tool" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "tool_name": "create-pull-request",
    "params": {
      "repository": "owner/repo",
      "title": "Auto-generated feature PR",
      "from_branch": "feature/auto-deploy",
      "to_branch": "main"
    }
  }'

​
Step 4: View Results

Check your GitHub repository and Slack channel:
  • ✅ PR created on GitHub
  • ✅ Slack message posted
  • ✅ All done in seconds!

​
What Just Happened?

  1. Retrieval - Your natural language query matched against 100+ available tools
  2. Ranking - Tools ranked by relevance (create PR scored 0.98, notify 0.92)
  3. Execution - Selected tools executed in order
  4. Results - Returned with status and output
This is the power of dynamic tool retrieval — your agent adapted to your request without any hardcoded configuration.

​
Next: Connect MCP Servers

Your agent is now connected to default MCP servers. To add more tools:
  1. Go to Dashboard → MCP Servers
  2. Browse the marketplace
  3. Click “Connect” on any server
  4. Your agent instantly gains access to its tools

GitHub

PRs, issues, deployments, CI/CD workflows

Slack

Messages, channels, user management

Jira

Issues, sprints, project management

​
Common Patterns

​
Pattern 1: Search + Execute

// Find the right tool first
const tools = await agent.retrieveTools({
  query: "Deploy to production using Terraform"
});

// Execute the top-ranked tool
const deployment = await agent.executeTool({
  toolName: tools[0].name,
  params: { environment: 'production' }
});

​
Pattern 2: Multi-Step Workflow

// Step 1: Create task
const task = await agent.executeTool({
  toolName: 'create-jira-issue',
  params: { title: 'Deploy v2.0' }
});

// Step 2: Update Slack
await agent.executeTool({
  toolName: 'send-slack-message',
  params: {
    channel: '#deployments',
    message: `New deployment task: ${task.id}`
  }
});

// Step 3: Trigger deployment
const deployment = await agent.executeTool({
  toolName: 'deploy-terraform',
  params: { version: '2.0' }
});

​
Pattern 3: Error Handling

try {
  const result = await agent.executeTool({
    toolName: 'create-pull-request',
    params: { /* ... */ }
  });
} catch (error) {
  if (error.code === 'TOOL_NOT_FOUND') {
    // Tool not connected, guide user to marketplace
  } else if (error.code === 'RATE_LIMITED') {
    // Retry with backoff
  } else {
    // Handle other errors
  }
}

​
Troubleshooting

Check that your API key is correctly set in the environment variable:
export AGENT_COREX_API_KEY="your_key"
echo $AGENT_COREX_API_KEY
Make sure the tool is connected in your MCP Servers dashboard. Use /retrieve_tools to see what’s available.
Free tier: 1,000 requests/month. Upgrade to Pro for higher limits (100,000/month).

​
Next Steps

Explore Core Concepts

Understand how dynamic retrieval works under the hood.

MCP Servers Guide

Learn how to connect and use MCP servers for developers.

Real Use Cases

See production examples for GitHub, DevOps, and workflows.

Full API Reference

Dive into detailed endpoint documentation.
Ready to build? Join our community Slack for questions and examples!