MCP
A2AThe AI Playground lets you test any action from your linked accounts using natural language.Documentation Index
Fetch the complete documentation index at: https://docs.stackone.com/llms.txt
Use this file to discover all available pages before exploring further.
Testing tool, not a production agent. The Playground is designed for quickly testing actions and exploring capabilities. For production use cases, build your own agent using our MCP Server, AI Toolset, or A2A Protocol. See Agent Protocols Comparison for guidance.
Access Required: Contact StackOne support if you don’t see the Playground in your dashboard.
Getting Started
Select accounts
Choose one or more linked accounts from the dropdown. Each account shows its provider and category.
Configure actions (optional)
Expand the Actions section to enable/disable specific actions before chatting.
Example Queries
The queries you can run depend on which accounts you’ve connected and what actions are available. Tailor your prompts to your agent’s use case:| Use Case | Example Query |
|---|---|
| Explore capabilities | ”What tools can I use with these accounts?” |
| List data | ”Show me the first 10 employees” |
| Get specific records | ”Find employee with email john@company.com” |
| Create records | ”Create a new contact with name Jane Doe” |
Model Selection
Click the model selector at the bottom right to switch between models:| Model | Recommendation |
|---|---|
| Anthropic Haiku | Recommended. Best balance of speed and tool calling accuracy for testing. |
| Anthropic Sonnet | Use for complex multi-step reasoning tasks. |
| GPT-4o | Alternative for comparison testing. |
StackOne Tool Calling Model (Beta)StackOne is fine-tuning a specialized model optimized for SaaS tool calling, at lower cost than leading models.Contact ai@stackone.com to join the beta.
Defender
The Playground includes a StackOne Defender toggle in the configuration sidebar that scans tool call responses for prompt injection attempts before the model sees them. Per-request override. The Playground toggle is a per-request override — it controls Defender for the Playground session only, regardless of what the project-level setting is. Use it to test how your agent behaves with Defender on or off without changing your project configuration. The toggle defaults to on in the Playground. Risk indicators on tool calls. When Defender is enabled, every tool call in the chat log displays a colored icon next to its status badge:| Icon | Meaning |
|---|---|
 | Low risk — response passed Defender’s checks |
 | Medium risk — flagged but not blocked |
 | High or critical risk — when Block High Risk Content is on in your project settings, this call is blocked from reaching the model |
| no icon | Defender did not scan this response (toggle off, size limit exceeded, or no defender metadata) |
Blocked status in the chat log and the model receives an error in place of the tool response. The agent can react like any other tool error — retry, skip, or surface to the user.
For the full Defender configuration reference — detection modes, thresholds, large-response behaviour, SFE, and SDK overrides — see the Defender guide.
Use Setup
The “Use Setup” button shows configuration for connecting StackOne’s MCP server to your own applications:
- URL:
https://api.stackone.com/mcp - Auth:
Basic <base64-encoded-api-key> - Header:
x-account-id: <your-account-id>
Troubleshooting
No accounts visible
No accounts visible
Ensure accounts are linked in the Accounts section and actions are enabled in your connector configuration.
Slow responses
Slow responses
Switch to Haiku (fastest) or reduce the number of selected accounts.
Tool not found
Tool not found
Verify the action is enabled in your integration configuration. Check Request Logs for details.
Related
Defender
Prompt-injection defense for tool call responses
Agent Protocols
Choose the right protocol for production
MCP Quickstart
Connect MCP clients