Skip to content
Neubird AI Docs Neubird AI Docs Neubird AI Docs

FAQ

Frequently Asked Questions

Section titled “Frequently Asked Questions”

Common questions about NeuBird MCP Server.

General

Section titled “General”

What is NeuBird MCP?

Section titled “What is NeuBird MCP?”

NeuBird MCP Server is a Model Context Protocol server that connects AI assistants (like Claude) to NeuBird’s Production Ops Agent platform for autonomous incident investigation and root cause analysis.

What can I do with NeuBird MCP?

Section titled “What can I do with NeuBird MCP?”
  • Investigate alerts automatically with AI-powered RCA
  • Get actionable corrective actions your coding agent can execute
  • Track MTTR and time saved metrics
  • Test investigation instructions before deployment
  • Connect multi-cloud environments (AWS, Azure, GCP)
  • Integrate with monitoring tools (Datadog, PagerDuty, etc.)

How much does it cost?

Section titled “How much does it cost?”

NeuBird MCP Server is free to use. You need a NeuBird account, which is a commercial product. Contact NeuBird for pricing.

Is my data secure?

Section titled “Is my data secure?”

Yes. NeuBird MCP uses read-only access to your cloud resources and telemetry. All credentials are stored securely by NeuBird and Neubird is SOC2 certified so you can trust them with security.

Installation & Setup

Section titled “Installation & Setup”

Which AI clients are supported?

Section titled “Which AI clients are supported?”
  • Claude Desktop - Anthropic’s native desktop app
  • Claude Code - Terminal-based AI assistant
  • Cursor - AI-powered code editor
  • GitHub Copilot - VS Code with MCP extension
  • Continue - VS Code extension with MCP support
  • Any MCP-compatible client

See Installation Guide for setup instructions for each client.

Do I need Node.js?

Section titled “Do I need Node.js?”

Yes, Node.js 20 or higher is required.

Can I use it without installing globally?

Section titled “Can I use it without installing globally?”

Yes, use npx mcp-server-neubird to run without installation.

What authentication methods are supported?

Section titled “What authentication methods are supported?”
  • Email/Password (local or remote) — credentials passed as environment variables or HTTP headers
  • Bearer Token (remote only) — a JWT token passed as an Authorization: Bearer <token> header, validated via Auth0

See Installation Guide for details.

Where are credentials stored?

Section titled “Where are credentials stored?”

In your MCP client configuration file:

  • Claude Desktop: ~/Library/Application Support/Claude/claude_desktop_config.json (macOS)
  • Claude Code:
    • ~/.claude.json (user scope - most common)
    • .mcp.json in project root (project scope - for teams)
    • Use /mcp command to see which file is being used

Features

Section titled “Features”

How many tools are available?

Section titled “How many tools are available?”

43 tools across 7 categories:

  • Projects (5)
  • Connections (10)
  • Investigations (11)
  • Instructions (7)
  • Analytics (4)
  • Discovery (2)
  • Help (1)

What makes NeuBird MCP unique?

Section titled “What makes NeuBird MCP unique?”

The instruction testing workflow - you can test investigation instructions on past incidents before deploying them, preventing bad instructions from affecting all investigations.

How long does an investigation take?

Section titled “How long does an investigation take?”
  • Typical: 30-90 seconds
  • First investigation: 5-10 minutes (initial connection sync)
  • Complex incidents: Up to 2 minutes

Can I investigate past incidents?

Section titled “Can I investigate past incidents?”

Yes! Use neubird_list_sessions to find past incidents, then neubird_continue_investigation to ask follow-up questions.

Section titled “Can I get a link to share an investigation?”

Yes! After listing investigations, you can request a direct link to view them in the NeuBird web interface:

Link for #2
Link for sessionID abc-123-def-456

This is useful for:

  • Sharing investigations with team members
  • Adding to incident tickets
  • Viewing in the full web interface
  • Bookmarking important investigations

Connections

Section titled “Connections”

Which cloud providers are supported?

Section titled “Which cloud providers are supported?”
  • AWS (CloudWatch, EC2, RDS, Lambda, ECS)
  • Azure (Azure Monitor, VMs, Functions)
  • GCP (Cloud Logging, Monitoring, Compute)

Which monitoring tools are supported?

Section titled “Which monitoring tools are supported?”
  • Datadog
  • PagerDuty
  • New Relic
  • And more…

Ask Claude: “What connection types are available?”

Do I need to connect all my cloud providers?

Section titled “Do I need to connect all my cloud providers?”

No, start with your primary cloud provider and main monitoring tool. Add others as needed.

How long does connection sync take?

Section titled “How long does connection sync take?”
  • First sync: 5-15 minutes depending on resource count
  • Subsequent syncs: Automatic, real-time
  • Large environments: Up to 15 minutes initial sync

What permissions do connections need?

Section titled “What permissions do connections need?”
  • AWS: ReadOnlyAccess IAM policy
  • Azure: Reader role
  • GCP: Viewer role
  • Datadog: Read-only API and app keys

Investigations

Section titled “Investigations”

How do I investigate an alert?

Section titled “How do I investigate an alert?”
1. Show me uninvestigated alerts
2. Investigate [alert-id or "the first one"]
3. Show me the RCA

Can I create investigations without alerts?

Section titled “Can I create investigations without alerts?”

Yes! Use manual investigations to analyze issues without existing alerts:

Investigate high latency in payment-service between 2pm-3pm EST on Jan 15.
Users reported checkout taking 30+ seconds.

Use cases:

  • Proactive analysis before alerts fire
  • Historical incident research
  • Testing “what-if” scenarios
  • Training and documentation

Tip: Be specific! Include service names, timeframes, and symptoms. Vague prompts like “something broke” won’t work well.

See Manual Investigations Guide for details.

What’s the difference between alert_id and session_uuid?

Section titled “What’s the difference between alert_id and session_uuid?”
  • alert_id: For NEW uninvestigated alerts (use with neubird_investigate_alert)
  • session_uuid: For EXISTING investigations (use with neubird_continue_investigation)

Can I ask follow-up questions?

Section titled “Can I ask follow-up questions?”

Yes! After getting an RCA, ask:

  • “Why did this happen?”
  • “Has this happened before?”
  • “How can we prevent this?”

What if the RCA is wrong?

Section titled “What if the RCA is wrong?”
  1. Ask clarifying follow-up questions
  2. Add instructions to guide future investigations
  3. Test instructions on past sessions
  4. Only deploy if RCA improves

Can I investigate multiple alerts at once?

Section titled “Can I investigate multiple alerts at once?”

Yes, in separate Claude conversations. Each conversation can handle one investigation at a time.

Instructions

Section titled “Instructions”

What are instructions?

Section titled “What are instructions?”

Instructions guide NeuBird’s investigation behavior:

  • FILTER: Reduce noise
  • SYSTEM: Provide context
  • GROUPING: Group related alerts
  • RCA: Investigation steps

How do I test an instruction?

Section titled “How do I test an instruction?”
1. Validate instruction
2. Apply to test session
3. Rerun session
4. Compare new vs old RCA
5. Add to project if improved

Can I modify instructions after creating them?

Section titled “Can I modify instructions after creating them?”

You can enable/disable them using neubird_update_project_instruction_status. To modify content, delete and recreate.

How many instructions should I have?

Section titled “How many instructions should I have?”

Start with:

  • 1-2 SYSTEM instructions (architecture context)
  • 1 FILTER instruction (priority threshold)
  • 2-3 RCA instructions (common incident types)

Add more as needed based on incident patterns.

Analytics

Section titled “Analytics”

What metrics can I track?

Section titled “What metrics can I track?”
  • MTTR (Mean Time To Resolution)
  • Time saved vs manual investigation
  • Investigation quality scores
  • Noise reduction from filtering
  • Incident patterns and trends

How is “time saved” calculated?

Section titled “How is “time saved” calculated?”

Compares estimated manual investigation time vs actual NeuBird investigation time.

Example:

  • Manual estimate: 45 minutes
  • NeuBird time: 60 seconds
  • Time saved: 44 minutes

What’s a good quality score?

Section titled “What’s a good quality score?”
  • 85-100: Excellent
  • 70-84: Good
  • <70: Needs improvement (add more instructions)

Troubleshooting

Section titled “Troubleshooting”

Tools not showing in Claude Desktop

Section titled “Tools not showing in Claude Desktop”
  1. Check config file syntax (JSON valid?)
  2. Verify file location
  3. Restart Claude Desktop completely
  4. Check logs: ~/Library/Logs/Claude/mcp*.log

Authentication failed

Section titled “Authentication failed”
  1. Verify email and password are correct
  2. Test login at NeuBird web UI
  3. No extra spaces in credentials
  4. For remote server: verify your access token is valid

Investigation taking too long

Section titled “Investigation taking too long”
  • First investigation: Normal (5-10 min sync)
  • Complex incidents: Normal (up to 2 min)
  • Check connection sync status

Connection won’t sync

Section titled “Connection won’t sync”
  1. Wait longer (first sync takes time)
  2. Verify credentials correct
  3. Check permissions (ReadOnly access)
  4. Check network connectivity

Best Practices

Section titled “Best Practices”
Section titled “What’s the recommended onboarding flow?”
  1. Install NeuBird MCP
  2. Create one project
  3. Add 1-2 connections (primary cloud + monitoring)
  4. Wait for sync
  5. Add basic instructions (SYSTEM + FILTER)
  6. Investigate first alert
  7. Refine instructions based on results

How often should I review instructions?

Section titled “How often should I review instructions?”
  • Weekly: Review investigation quality scores
  • Bi-weekly: Update instructions based on patterns
  • Monthly: Clean up unused instructions

Should I create separate projects for each environment?

Section titled “Should I create separate projects for each environment?”

Yes, recommended:

  • Production/Staging/Development projects
  • Application specific projects
  • Team specific projects
  • Alert specific projects

Each can have different connections and instructions. You can use the same connections in multiple projects and by using investigation instructions you can control which alerts gets investigated in each project. This makes it possible to have specialized instructions in separate projects.

How do I switch between projects?

Section titled “How do I switch between projects?”

Use neubird_set_default_project to switch your active project context:

Switch to my Staging project

Or use the project UUID directly:

neubird_set_default_project(project_uuid="abc-123...")

Benefits:

  • No need to specify project_uuid in every command
  • Quick context switching between projects
  • Natural language support (use project names)
  • Default persists for entire MCP session

Example workflow:

# Set Production as default
Switch to Production project


# All commands now use Production
List my project's connections
Show recent investigations


# Switch to Staging for testing
Switch to Staging project


# Now everything uses Staging
Test the new instruction

The default project is shown with a ⭐ emoji when you list projects.

What’s the fastest way to get started?

Section titled “What’s the fastest way to get started?”

Follow the Quick Start Guide - takes 5 minutes.

Advanced

Section titled “Advanced”

Can I integrate with CI/CD?

Section titled “Can I integrate with CI/CD?”

Yes, you can trigger investigations from deployment pipelines and extract corrective actions programmatically.

Can I export investigation results?

Section titled “Can I export investigation results?”

Yes, RCA results are returned as formatted text that you can copy, export, or integrate with ticketing systems.

What’s the API rate limit?

Section titled “What’s the API rate limit?”

NeuBird MCP respects NeuBird API rate limits. For high-volume usage, contact NeuBird support.

Support

Section titled “Support”

Where can I get help?

Section titled “Where can I get help?”
  1. Documentation: Full docs
  2. Inline help: Ask Claude “How do I…” using the guidance system
  1. Support: Email support@neubird.ai

How do I report a bug?

Section titled “How do I report a bug?”

Email support@neubird.ai with:

  • Error message
  • Steps to reproduce
  • MCP server version
  • Logs (if available)

Where’s the source code?

Section titled “Where’s the source code?”

The MCP server is distributed as an npm package. The code is currently private.

How do I request a feature?

Section titled “How do I request a feature?”

Contact support@neubird.ai or your NeuBird account representative.

See Also

Section titled “See Also”