- Démarrage
- Meilleures pratiques
- Modélisation de l'organisation dans Orchestrator
- Meilleures pratiques d'automatisation
- Optimisation de l'infrastructure Unattended à l'aide de modèles de machine
- Organisation des ressources avec des balises
- Exportation des grilles dans l'arrière-plan
- Appliquer la gouvernance de la connexion Integration Service au niveau de l'utilisateur
- Locataire
- À propos du contexte du locataire
- Recherche de ressources dans un locataire
- Gestion des Robots
- Connexion des Robots à Orchestrator
- Enregistrement des identifiants du Robot dans CyberArk
- Stockage des mots de passe de l’Unattended Robot dans Azure Key Vault (lecture seule)
- Stockage des informations d’identification de l’Unattended Robot dans HashiCorp Vault (lecture seule)
- Stockage des informations d'identification du robot Unattended dans AWS Secrets Manager (lecture seule)
- Suppression des sessions Unattended déconnectées et qui ne répondent pas
- Authentification du Robot
- Authentification du Robot avec les informations d'identification du client
- Configurer les capacités d’automatisation
- Solutions
- Audit
- Paramètres
- Registre
- Cloud Robots
- Vue d'ensemble des robots Cloud
- Exécution d'automatisations Unattended à l'aide de Cloud Robots - VM
- Téléchargement de votre propre image
- Réutilisation des images de machines personnalisées (pour les pools manuels)
- Réinitialisation des informations d'identification d'une machine (pour les pools manuels)
- Surveillance
- Mises à jour de sécurité
- Demander un essai
- Questions fréquemment posées
- Configuration du VPN pour les robots du cloud
- Configurer une connexion ExpressRoute
- Diffusion en direct et contrôle à distance
- Robots Automation Suite
- Contexte des dossiers
- Processus (Processes)
- Tâches (Jobs)
- Apps
- Déclencheurs (Triggers)
- Journaux (Logs)
- Surveillance
- Index
- Files d'attente (Queues)
- Actifs
- À propos des actifs
- Gestion des actifs dans Orchestrator
- Gestion des actifs dans Studio
- Stockage des ressources dans Azure Key Vault (lecture seule)
- Stockage des ressources dans HashiCorp Vault (lecture seule)
- Stockage des ressources dans AWS Secrets Manager (lecture seule)
- Stocker des ressources dans Google Secret Manager (lecture seule)
- Connexions
- Règles métier
- Compartiments de stockage
- Serveurs MCP
- Tests d'Orchestrator
- Service de catalogue de ressources
- Intégrations
- Résolution des problèmes
Test a UiPath MCP Server directly, outside any AI client, using the UiPath MCP Inspector, the open-source MCP Inspector, cURL, or Postman.
Before connecting an MCP Server to an AI client, test the connection and tool behavior directly. Testing outside an AI client lets you isolate whether an issue originates in the MCP Server, its authentication, or the AI client.
Testing with the UiPath MCP Inspector
The UiPath MCP Inspector runs inside Orchestrator and requires no separate setup.
Access the Inspector
-
À la page Serveurs MCP, sélectionnez un serveur MCP.
-
Select the Inspector icon.
Call a tool
-
Select a tool from the list.
-
Enter the tool's input parameters.
-
Select Invoke to inspect the result.
Testing with the MCP Inspector (CLI)
The MCP Inspector is an open-source tool that connects to any MCP Server over Streamable HTTP.
Prérequis
- Node.js is installed to run the
npxcommand. - The MCP Server URL, available from the MCP Servers page in Orchestrator.
- A bearer token. For the simplest option, use a personal access token (PAT). For other supported token types, check MCP Server authentication.
-
From a terminal, run the following command. It is the same on macOS, Linux, and Windows (PowerShell 7 or Git Bash):
npx @modelcontextprotocol/inspector@0.22.0npx @modelcontextprotocol/inspector@0.22.0The command opens the Inspector web UI in your browser.
-
Enter the MCP Server URL.
-
Select Streamable HTTP as the transport type.
-
In the Authentication section, enter the token in the format
Bearer <token>. -
Select Connect.
Résultat
Once connected, you can:
- List tools to verify all your tools appear with correct JSON schemas.
- Call tools to execute tools with test inputs and inspect responses.
- Watch the session to see the
initialize→initialized→requestlifecycle in real time. - Validate schemas to catch type mismatches before they hit an LLM.
Testing with cURL
Use cURL to send raw JSON-RPC requests and inspect the full request/response cycle, including headers.
Prérequis
- A bearer token. Check MCP Server authentication for supported methods.
- The MCP Server URL, available from the MCP Servers page in Orchestrator.
-
Set the token and server URL as environment variables:
TOKEN="your-bearer-token" SERVER_URL="https://cloud.uipath.com/{org}/{tenant}/agenthub_/mcp/{folder}/{slug}"TOKEN="your-bearer-token" SERVER_URL="https://cloud.uipath.com/{org}/{tenant}/agenthub_/mcp/{folder}/{slug}" -
Initialize the MCP session:
curl -v -X POST "$SERVER_URL" \ -H "Authorization: Bearer $TOKEN" \ -H "Content-Type: application/json" \ -H "Accept: application/json, text/event-stream" \ -d '{ "jsonrpc": "2.0", "id": 1, "method": "initialize", "params": { "protocolVersion": "2025-06-18", "capabilities": {}, "clientInfo": {"name": "curl-test", "version": "1.0"} } }'curl -v -X POST "$SERVER_URL" \ -H "Authorization: Bearer $TOKEN" \ -H "Content-Type: application/json" \ -H "Accept: application/json, text/event-stream" \ -d '{ "jsonrpc": "2.0", "id": 1, "method": "initialize", "params": { "protocolVersion": "2025-06-18", "capabilities": {}, "clientInfo": {"name": "curl-test", "version": "1.0"} } }'Use the
-vflag to see response headers, and save themcp-session-idheader value. Every subsequent request needs it.Remarque :UiPath MCP Servers support protocol version
2025-06-18. Sending2025-11-25(the latest MCP spec version) may work, since the server responds with the version it supports, but use2025-06-18to avoid unsupported behavior. -
Send the initialized notification:
curl -X POST "$SERVER_URL" \ -H "Authorization: Bearer $TOKEN" \ -H "Content-Type: application/json" \ -H "mcp-session-id: $SESSION_ID" \ -d '{"jsonrpc": "2.0", "method": "notifications/initialized"}'curl -X POST "$SERVER_URL" \ -H "Authorization: Bearer $TOKEN" \ -H "Content-Type: application/json" \ -H "mcp-session-id: $SESSION_ID" \ -d '{"jsonrpc": "2.0", "method": "notifications/initialized"}'The server responds with
202 Acceptedand no body. The server requires this notification before it processes tool requests. -
List the available tools:
curl -X POST "$SERVER_URL" \ -H "Authorization: Bearer $TOKEN" \ -H "mcp-session-id: $SESSION_ID" \ -H "Content-Type: application/json" \ -d '{ "jsonrpc": "2.0", "id": 2, "method": "tools/list" }'curl -X POST "$SERVER_URL" \ -H "Authorization: Bearer $TOKEN" \ -H "mcp-session-id: $SESSION_ID" \ -H "Content-Type: application/json" \ -d '{ "jsonrpc": "2.0", "id": 2, "method": "tools/list" }' -
Call a tool:
curl -X POST "$SERVER_URL" \ -H "Authorization: Bearer $TOKEN" \ -H "mcp-session-id: $SESSION_ID" \ -H "Content-Type: application/json" \ -d '{ "jsonrpc": "2.0", "id": 3, "method": "tools/call", "params": { "name": "add", "arguments": {"a": 5, "b": 3} } }'curl -X POST "$SERVER_URL" \ -H "Authorization: Bearer $TOKEN" \ -H "mcp-session-id: $SESSION_ID" \ -H "Content-Type: application/json" \ -d '{ "jsonrpc": "2.0", "id": 3, "method": "tools/call", "params": { "name": "add", "arguments": {"a": 5, "b": 3} } }' -
Close the session:
curl -X DELETE "$SERVER_URL" \ -H "Authorization: Bearer $TOKEN" \ -H "mcp-session-id: $SESSION_ID"curl -X DELETE "$SERVER_URL" \ -H "Authorization: Bearer $TOKEN" \ -H "mcp-session-id: $SESSION_ID"
Testing with Postman
Postman supports MCP requests natively.
- Create an MCP request in your collection.
- Select HTTP as the transport, for Streamable HTTP.
- Enter the MCP Server URL.
- In the Authorization tab, configure Bearer Token authentication with your token.
- Select Connect.
Postman handles the session lifecycle, including initialize and notifications/initialized, automatically.
To send raw JSON-RPC requests manually instead, for example to control each request and response individually, use Postman as a plain REST client and follow the same request flow as Testing with cURL.