syncause-debug-mcp
# Syncause Debug MCP Server
**Without Syncause**, coding agents often guess and only fix symptoms, so you may need to prompt the agent multiple times to fix the root cause.
**With Syncause**, agents can fix the root cause in one go. Syncause captures **runtime facts** (stack traces, logs, request/response, function arguments, and key variable values) and makes it queryable by AI agents via **MCP**—so the agent debugs with evidence instead of guessing.
**Use it with:**
- [Syncause IDE Extension](https://open-vsx.org/extension/syncause/ai-debugger) — capture runtime data in your IDE, then query it via MCP.
- [Syncause Debugger Skill](https://github.com/syncause/debug-skill) — a prompt/workflow layer that calls this MCP server.
## Installation
> [!NOTE]
> **API Key Recommended**: Get a free API key at [syn-cause.com/dashboard](https://syn-cause.com/dashboard) for higher rate limits.
Common MCP Configuration works in most of the tools:
```json
{
"mcpServers": {
"syncause-debug-mcp": {
"command": "npx",
"args": ["-y", "@syncause/debug-mcp@latest"],
"env": { "API_KEY": "YOUR_API_KEY" }
}
}
}
```
### Configure in your client
<details>
<summary><b>Install in Cursor</b></summary>
Edit `.cursor/mcp.json` (project) or `~/.cursor/mcp.json` (global):
```json
{
"mcpServers": {
"syncause-debug-mcp": {
"command": "npx",
"args": ["-y", "@syncause/debug-mcp@latest"],
"env": { "API_KEY": "YOUR_API_KEY" }
}
}
}
```
</details>
<details>
<summary><b>Install in VS Code</b></summary>
Edit `.vscode/settings.json`:
```json
{
"mcp.servers": {
"syncause-debug-mcp": {
"command": "npx",
"args": ["-y", "@syncause/debug-mcp@latest"],
"env": { "API_KEY": "YOUR_API_KEY" }
}
}
}
```
</details>
<details>
<summary><b>Install in Antigravity</b></summary>
Edit `~/.gemini/antigravity/mcp_config.json` (Global):
```json
{
"mcp.servers": {
"syncause-debug-mcp": {
"command": "npx",
"args": ["-y", "@syncause/debug-mcp@latest"],
"env": { "API_KEY": "YOUR_API_KEY" }
}
}
}
```
</details>
<details>
<summary><b>Install in Claude Code</b></summary>
Run this command (local stdio server):
```sh
API_KEY=YOUR_API_KEY claude mcp add syncause-debug-mcp -- npx -y @syncause/debug-mcp@latest
```
</details>
<details>
<summary><b>Install in OpenAI Codex (CLI)</b></summary>
```sh
API_KEY=YOUR_API_KEY codex mcp add syncause-debug-mcp --command "npx -y @syncause/debug-mcp@latest"
```
</details>
<details>
<summary><b>Install in Gemini CLI</b></summary>
```sh
API_KEY=YOUR_API_KEY gemini mcp add syncause-debug-mcp -- npx -y @syncause/debug-mcp@latest
```
</details>
<details>
<summary><b>Install in Windsurf</b></summary>
Edit `~/.codeium/windsurf/mcp_config.json`:
```json
{
"mcpServers": {
"syncause-debug-mcp": {
"command": "npx",
"args": ["-y", "@syncause/debug-mcp@latest"],
"env": { "API_KEY": "YOUR_API_KEY" }
}
}
}
```
</details>
<details>
<summary><b>Install in Goose</b></summary>
Edit `~/.config/goose/config.yaml`:
```yaml
extensions:
syncause-debug-mcp:
type: stdio
name: Syncause Debug MCP
enabled: true
timeout: 300
cmd: npx
args: ["-y", "@syncause/debug-mcp@latest"]
envs:
API_KEY: "YOUR_API_KEY"
```
</details>
<details>
<summary><b>Install in Opencode</b></summary>
Add this to your Opencode config:
```json
{
"mcp": {
"syncause-debug-mcp": {
"type": "local",
"command": ["npx", "-y", "@syncause/debug-mcp@latest"],
"env": { "API_KEY": "YOUR_API_KEY" },
"enabled": true
}
}
}
```
</details>
## Troubleshooting
- **`npx: command not found`** → install Node.js (or ensure your shell can find it).
- **Server won’t start in your client** → try running the MCP server directly in a terminal first:
```bash
npx -y @syncause/debug-mcp@latest
```
- **Env not applied** → many clients require restarting after config changes.
## Security Notes
- Your `API_KEY` is sensitive. Avoid committing MCP config files to Git.
- Prefer project-level configs only when you need per-repo isolation.
TDQS
Scored across 6 tools
Each tool targets a distinct level of the tracing workflow: discovery (search_debug_traces), request-level report (get_trace_insight), method-level drill-down (inspect_method_snapshot), comparison (diff_trace_execution), and project config (setup_project/get_project_list). The only mild overlap is between search_debug_traces and get_trace_insight, since both surface logs/errors, but the descriptions clarify search is the entry point and insight is a full lifecycle report.
All names follow a snake_case verb_noun pattern (get_trace_insight, search_debug_traces, inspect_method_snapshot, diff_trace_execution, setup_project, get_project_list). Minor deviation: 'diff' and 'setup' are less standard verbs than get/search/inspect, and project tools split verbs (get vs setup) for related concepts, but overall the convention is predictable.
Six tools is well-scoped for a trace-debugging server: two for project setup/discovery and four covering the search-to-diagnosis-to-comparison flow. No tool appears redundant or filler.
The core lifecycle (discover project, search traces, inspect request and method detail, diff two requests) is fully covered. Minor gaps exist around project config lifecycle (no removal/reset of an initialized project) and no way to list or browse recent traces without a search query, but agents can work around these.