INTEGRATIONS
● Connections: 27 active
·
1,247 calls today
·
100% authenticated
The integration layer your agents need - without the integration project.
1,000+
MCP INTEGRATIONS
200+
BUILT-IN TOOLS
1-click
OAUTH SETUP
Per-agent
PERMISSIONS SCOPED
THE TOOL SURFACE
Built-in Tools
Tools that ship with the platform. Document generation (PDF, Excel, Slides). Image and video generation. Lead capture and CRM forwarding. Scheduling and calendar. Email send. 200+ tools, zero setup. Toggle on per agent.
MCP Integrations
Pre-configured Model Context Protocol servers. Salesforce, HubSpot, Slack, GitHub, Notion, Stripe, PostgreSQL, Mailchimp, and 1,000+ more. One-click install. OAuth or API key auth. Updates and security patches managed by us.
Custom Tools
When something doesn't exist yet, write it. In-platform Python editor. Vibe-code via natural language ("build me a tool that verifies SAT CFDI tax IDs"). Test, version, deploy - all inside the same workspace.
BUILT-IN
All
Document
Media
Capture
Schedule
Communication
Productivity
📄 PDF Generator
High-quality PDF creation from agent output.
Enable
📊 Excel Generator
Spreadsheets with formulas and formatting.
Enable
🎬 Slides Generator
Presentations from agent content.
Enable
🖼 Image Generator
Text-to-image generation.
Enable
🎥 Video Generator
Text-to-video generation.
Enable
💼 Office Generators
DOCX, XLSX, PPTX bundle.
Enable
🎯 Lead Capture
Inbound lead form with CRM forwarding.
Enable
Google Sheets
Read/write sheets directly.
Enable
Google Calendar
Create, modify, delete events.
Enable
📞 Calendly
Manage and display bookings.
Enable
Mailchimp Send
Marketing email send + lists.
Enable
🔗 Web Tool
Fetch and parse web pages.
Enable
MCP MARKETPLACE
Featured
CRM
Communication
Cloud & DB
Productivity
E-commerce
Dev Tools
Marketing
Finance & Data
Salesforce
HubSpot
Slack
Mailchimp
PostgreSQL
MongoDB
Astra DB
Notion
Google Sheets
Stripe
GitHub
GitLab
Playwright
ActiveCampaign
AirTable
Alphavantage
Yahoo Finance
Google Calendar
Always up to date
Servers are managed centrally. Security patches, API updates, breaking change migrations - handled. You don't maintain integrations.
Auth managed
OAuth flows, API key vaults, token refresh - handled per workspace. Credentials never reach individual agents.
Open source
AgentX maintains an open-source MCP server collection on GitHub. Want a specific integration? Request it or build it.
CUSTOM
Custom Function Editor
verify_sat_cfdi.py
v2
✓ Tested
TOOL NAME
verify_sat_cfdi
DESCRIPTION
Verifies a Mexican CFDI tax ID via SAT SOAP service with the verification URL and parameters.
PARAMETERS
verification_url (string, required)
TESTS
✓ Valid CFDI ID
✓ Invalid CFDI ID
✓ Network timeout
✓ Malformed URL
def verify_sat_cfdi(verification_url):
"""Verifies a Mexican CFDI (Comprobante Fiscal
Digital por Internet) by calling SAT's SOAP service
with the verification URL, returns full status."""
import re, json, requests
from urllib.parse import urlparse, parse_qs
if not verification_url:
return json.dumps({
"error": "verification_url is required",
"success": False
})
# Parse URL to extract parameters
parsed = urlparse(verification_url)
params = parse_qs(parsed.query)
# ... (rest of implementation)
Save
Run test
Make public
Add to agent
✓
Write Python directly or describe what you need in plain English.
✓
Test cases auto-generated, runnable from the editor.
✓
Version-controlled and shareable across agents in a workspace.
✓
Custom tools follow the same auth, permission, and audit model as built-ins.
UNDER THE HOOD
STAGE 1
Agent decides
Agent calls a tool. Permission check fires.
→
SECURITY · STAGE 2
Permission gate
Workspace policy + per-agent allow-list.
→
SECURITY · STAGE 3
Auth layer
OAuth/API key from workspace vault.
→
STAGE 4
Tool execution
Built-in, MCP, or custom function.
→
STAGE 5
Response handling
Return to agent + log to audit trail.
→
STAGE 6
Trace event
Recorded in observability layer.
⚠ Security perimeter spans stages 2–3
Per-agent permissions
An agent can only call tools it's explicitly allowed to call. A customer support agent can't suddenly query your production database. Permissions defined at agent creation, modifiable per environment.
Auth at the workspace, not the agent
OAuth tokens and API keys live in a workspace vault - encrypted at rest, never exposed to the agent context. Tool calls happen on the agent's behalf, not with the agent's credentials.
Every call logged
Tool name, parameters, response, latency, cost. Searchable for 90 days (longer on Enterprise). Exportable. Audit-ready.
Fail-safe by default
Rate-limited tools queue with backoff. Failed tools route to fallback or human queue. No tool failure causes silent agent breakdown - every error is observable.
BUILT FOR REAL USERS
Workspace credential vault
All API keys, OAuth tokens, and secrets encrypted at rest. Never visible in agent context. Never logged in plain text.
Secure execution sandbox
Custom Python tools run in isolated sandboxes. No filesystem access outside designated paths. No network calls outside whitelisted endpoints (configurable per workspace).
Tool-call audit trail
Every tool call logged with full context. SOC 2 audit ready. Export to your SIEM via OTel or syslog.
On-prem MCP support
Run MCP servers inside your perimeter. AgentX agents connect to on-prem servers without your data leaving your environment.
GET STARTED
