DevOps-OS MCP: Environment Variables Reference#

Complete reference for all environment variables that control DevOps-OS MCP server behavior and features.

Quick Reference#

Transport & Server Configuration#

DEVOPS_OS_TRANSPORT#

Type: String
Default: stdio
Valid Values: stdio, sse, streamable-http

Controls how the MCP server communicates with clients.

Options:

• stdio (Standard Input/Output)

  • Use for: Claude Desktop (local)
  • No network exposure
  • Simplest setup

• streamable-http (HTTP with streaming)

  • Use for: Remote deployments, ChatGPT, Docker
  • Network-accessible
  • Requires auth for remote profile

• sse (Server-Sent Events)

  • Alternative HTTP transport
  • Use for: Specific client requirements

Example:

# Local (Claude Desktop)
export DEVOPS_OS_TRANSPORT=stdio

# Remote (ChatGPT, HTTP)
export DEVOPS_OS_TRANSPORT=streamable-http

DEVOPS_OS_PROFILE#

Type: String
Default: local
Valid Values: local, remote

Deployment profile that controls authentication and network binding.

Options:

• local (Development)

  • No authentication
  • Loopback only by default
  • Fast setup for local testing

• remote (Production)

  • Requires JWT authentication
  • Can bind to all interfaces
  • Suitable for shared deployments

Example:

# Development
export DEVOPS_OS_PROFILE=local

# Production (requires JWT configuration)
export DEVOPS_OS_PROFILE=remote
export DEVOPS_OS_JWT_ISSUER=https://auth.example.com/
export DEVOPS_OS_JWT_AUDIENCE=devops-os-service

Related Variables:

• DEVOPS_OS_JWT_ISSUER
• DEVOPS_OS_JWT_AUDIENCE
• DEVOPS_OS_JWT_JWKS_URL
• DEVOPS_OS_JWT_ALGORITHMS

HTTP Server Configuration#

Only applies when using DEVOPS_OS_TRANSPORT=streamable-http

DEVOPS_OS_HOST#

Type: String
Default: 127.0.0.1

IP address the HTTP server binds to.

Use Cases:

• 127.0.0.1 - Localhost only (most secure, default)
• 0.0.0.0 - All interfaces (required for Docker/remote access)
• 192.168.1.100 - Specific network interface

Example:

# Bind to all interfaces (for Docker, remote deployment)
export DEVOPS_OS_HOST=0.0.0.0

# Bind to specific interface
export DEVOPS_OS_HOST=192.168.1.100

# Default: localhost only
export DEVOPS_OS_HOST=127.0.0.1

Security Note: Only use 0.0.0.0 with DEVOPS_OS_PROFILE=remote and proper JWT authentication in production.

DEVOPS_OS_PORT#

Type: Integer
Default: 8000
Valid Range: 1-65535

Port the HTTP server listens on.

Common Values:

• 8000 - Default, suitable for development
• 8080 - Alternative HTTP port
• 9000 - Higher port number
• 443 - HTTPS (if running behind proxy)

Example:

# Standard port
export DEVOPS_OS_PORT=8000

# Non-standard (if 8000 is in use)
export DEVOPS_OS_PORT=9000

# With docker compose
export DEVOPS_OS_PORT=8001

Constraint: When changing ports, ensure firewall rules and proxy configurations are updated accordingly.

DEVOPS_OS_MCP_ENDPOINT#

Type: String
Default: /mcp

HTTP path where the MCP server handles requests.

Use Cases:

• /mcp - Default, simple routing
• /tools/mcp - Nested under tools prefix
• /devops-os - Application-specific path
• /api/mcp - RESTful convention

Example:

# Default
export DEVOPS_OS_MCP_ENDPOINT=/mcp

# Behind a proxy with path prefix
export DEVOPS_OS_MCP_ENDPOINT=/devops-os/mcp

# Nested under tools
export DEVOPS_OS_MCP_ENDPOINT=/tools/mcp

Note: If running behind a reverse proxy, update your proxy configuration to match this path.

Logging Configuration#

DEVOPS_OS_LOG_LEVEL#

Type: String
Default: INFO
Valid Values: DEBUG, INFO, WARNING, ERROR, CRITICAL

Controls verbosity of server logs (Python logging levels).

Log Levels (from most to least verbose):

• DEBUG - Detailed diagnostic information, variable states, function calls

  • Use for: Troubleshooting issues
  • May include sensitive data

• INFO - General informational messages

  • Use for: Normal operation, monitoring
  • Default level

• WARNING - Warning messages about potential issues

  • Use for: Production (reduced noise)

• ERROR - Error messages only

  • Use for: Critical production systems

• CRITICAL - Only critical failures

Example:

# Development: Maximum detail
export DEVOPS_OS_LOG_LEVEL=DEBUG

# Production: Normal monitoring
export DEVOPS_OS_LOG_LEVEL=INFO

# Production: Warnings only
export DEVOPS_OS_LOG_LEVEL=WARNING

Viewing Logs:

# Start server and see logs
export DEVOPS_OS_LOG_LEVEL=DEBUG
python -m mcp_server.server

# Or tail logs if running in background
docker logs -f devops-os-mcp

Performance & Resource Tuning#

DEVOPS_OS_REQUEST_SIZE_MB#

Type: Integer
Default: 10
Valid Range: 1-1024 MB

Maximum size of incoming request body in megabytes.

Use Cases:

• 10 - Small/normal prompts (default)
• 50 - Larger prompts with detailed context
• 100 - Very large inputs or batch operations

Example:

# Default: 10 MB
export DEVOPS_OS_REQUEST_SIZE_MB=10

# For large prompt inputs
export DEVOPS_OS_REQUEST_SIZE_MB=50

# For very large inputs (if needed)
export DEVOPS_OS_REQUEST_SIZE_MB=100

When to Increase:

• Users report “request body too large” errors
• Handling large context windows from AI assistants
• Processing complex multi-config requests

When to Keep Small:

• Limited server memory
• Security constraints on input sizes
• Normal use cases

DEVOPS_OS_RESPONSE_SIZE_MB#

Type: Integer
Default: 50
Valid Range: 1-1024 MB

Maximum size of outgoing response body in megabytes.

Use Cases:

• 50 - Standard responses (default)
• 100 - Large generated configurations
• 200 - Multiple large artifacts

Example:

# Default: 50 MB
export DEVOPS_OS_RESPONSE_SIZE_MB=50

# For large generated configs
export DEVOPS_OS_RESPONSE_SIZE_MB=100

# For very large responses
export DEVOPS_OS_RESPONSE_SIZE_MB=200

When to Increase:

• Generating multiple large Kubernetes manifests
• Creating comprehensive SRE dashboards
• Complex multi-service deployments

Typical Response Sizes:

• GitHub Actions workflow: 5-20 KB
• Kubernetes manifests: 10-100 KB
• Prometheus rules: 5-50 KB
• Multiple artifacts: 50-500 KB

DEVOPS_OS_EXECUTION_TIMEOUT#

Type: Integer
Default: 30 seconds
Valid Range: 1-600 seconds

Maximum time a single tool can execute before timing out.

Use Cases:

• 30 - Normal tool execution (default)
• 60 - Slower infrastructure
• 120 - Very slow systems
• 300 - Extended timeout for complex operations

Example:

# Default: 30 seconds
export DEVOPS_OS_EXECUTION_TIMEOUT=30

# For slower systems
export DEVOPS_OS_EXECUTION_TIMEOUT=60

# For complex operations
export DEVOPS_OS_EXECUTION_TIMEOUT=120

Common Tool Execution Times:

• GitHub Actions: 2-5 seconds
• Kubernetes configs: 1-3 seconds
• Jenkins pipelines: 3-8 seconds
• SRE configs: 2-5 seconds

When to Increase:

• “Tool execution timeout” errors occur
• Running on slow/limited hardware
• Complex custom scaffolding

When to Keep Low:

• Strict SLA requirements
• Limited server resources
• Quick response needed

DEVOPS_OS_MAX_CONCURRENT_CALLS#

Type: Integer
Default: 10
Valid Range: 1-100

Limits the maximum number of tool executions that can run concurrently. Prevents resource exhaustion and enforces SLA compliance in production.

Use Cases:

• 1 - Strictly sequential execution (debugging, very limited resources)
• 5 - Conservative limit (small servers, cost control)
• 10 - Default, balanced for most scenarios
• 20 - High-capacity servers
• 50+ - Large-scale deployments with ample resources

Example:

# Default: 10 concurrent calls
export DEVOPS_OS_MAX_CONCURRENT_CALLS=10

# Limit to 5 for small server
export DEVOPS_OS_MAX_CONCURRENT_CALLS=5

# Allow 20 for high-capacity deployment
export DEVOPS_OS_MAX_CONCURRENT_CALLS=20

# Strict sequential (no concurrency)
export DEVOPS_OS_MAX_CONCURRENT_CALLS=1

When to Lower (1-5):

• Limited server CPU/memory
• Cost-constrained deployments
• Strict resource quotas
• Shared infrastructure
• Testing/debugging

When to Keep Default (10):

• Standard deployments
• Most production scenarios
• Balanced performance

When to Increase (20+):

• High-capacity cloud deployments (e.g., Kubernetes)
• High traffic expected
• Ample CPU/memory available
• Auto-scaling infrastructure
• Premium tier deployments

Production Safety:

• Set based on available server resources
• Monitor CPU/memory under load
• Start conservative, increase if needed
• Each tool uses ~50-200MB during execution
• Consider: max_concurrent_calls × avg_tool_memory ≤ available_memory

Example Resource Calculation:

Server: 4 CPU, 8GB RAM
Average tool memory: 100MB
Recommended max_concurrent_calls = (8GB × 0.75) / 100MB ≈ 60
But also consider CPU: 4 cores suggests 8-16 concurrent is reasonable
Conservative choice: 10
Aggressive choice: 20

Authentication (JWT)#

Only applies when using DEVOPS_OS_PROFILE=remote

DEVOPS_OS_JWT_ISSUER#

Type: String
Default: None (required for remote profile)

URL of the JWT token issuer.

Common Issuers:

• https://auth.example.com/ - Custom auth server
• https://accounts.google.com/ - Google OAuth
• https://login.microsoftonline.com/ - Azure AD
• https://github.com/login/oauth/ - GitHub OAuth

Example:

# Custom auth server
export DEVOPS_OS_JWT_ISSUER=https://auth.company.com/

# Google
export DEVOPS_OS_JWT_ISSUER=https://accounts.google.com/

# Azure AD
export DEVOPS_OS_JWT_ISSUER=https://login.microsoftonline.com/YOUR_TENANT_ID/

Related Variables:

• DEVOPS_OS_JWT_AUDIENCE - Must match token’s ‘aud’ claim
• DEVOPS_OS_JWT_JWKS_URL - Auto-derived if not set

DEVOPS_OS_JWT_AUDIENCE#

Type: String
Default: None (required for remote profile)

Expected JWT audience claim value. The token’s aud claim must match this value.

Format: Usually a resource identifier or application name.

Example:

# Application identifier
export DEVOPS_OS_JWT_AUDIENCE=devops-os-service

# API resource
export DEVOPS_OS_JWT_AUDIENCE=api://devops-os

# Custom identifier
export DEVOPS_OS_JWT_AUDIENCE=urn:devops-os:mcp

Verification: Tokens without matching audience will be rejected.

DEVOPS_OS_JWT_JWKS_URL#

Type: String
Default: Auto-derived from DEVOPS_OS_JWT_ISSUER

URL of the JWKS (JSON Web Key Set) endpoint for token validation.

Auto-Derivation: If not set, derived by appending /.well-known/jwks.json to the issuer URL.

Example:

• Issuer: https://auth.example.com/
• Auto-derived: https://auth.example.com/.well-known/jwks.json

Example:

# Let it auto-derive (recommended)
export DEVOPS_OS_JWT_ISSUER=https://auth.example.com/
# Automatically: JWKS URL = https://auth.example.com/.well-known/jwks.json

# Explicit JWKS URL (if auto-derivation doesn't work)
export DEVOPS_OS_JWT_JWKS_URL=https://auth.example.com/oauth/jwks

When to Set Explicitly:

• Non-standard JWKS endpoint path
• Different domain for JWKS vs issuer
• Proxy or custom configuration

DEVOPS_OS_JWT_ALGORITHMS#

Type: String (comma-separated)
Default: RS256,ES256

Allowed JWT signing algorithms.

Common Algorithms:

• RS256 - RSA with SHA-256 (most common)
• ES256 - ECDSA with SHA-256
• HS256 - HMAC with SHA-256
• RS512 - RSA with SHA-512

Example:

# Default: Both RSA and ECDSA
export DEVOPS_OS_JWT_ALGORITHMS=RS256,ES256

# RSA only (most common)
export DEVOPS_OS_JWT_ALGORITHMS=RS256

# Multiple algorithms
export DEVOPS_OS_JWT_ALGORITHMS=RS256,RS512,ES256

# HMAC (less common, not recommended for server signing)
export DEVOPS_OS_JWT_ALGORITHMS=HS256

Security Note: Restrict to algorithms your auth provider actually uses to prevent downgrade attacks.

Prompt Improvement Suggestions#

DEVOPS_OS_ENABLE_SUGGESTIONS#

Type: Boolean
Default: true
Valid Values: true, false, 1, 0, yes, no

Enable or disable prompt improvement suggestion feature.

Example:

# Enable suggestions (default)
export DEVOPS_OS_ENABLE_SUGGESTIONS=true

# Disable suggestions
export DEVOPS_OS_ENABLE_SUGGESTIONS=false

# Alternative formats
export DEVOPS_OS_ENABLE_SUGGESTIONS=1      # enabled
export DEVOPS_OS_ENABLE_SUGGESTIONS=0      # disabled

What Suggestions Do:

• Analyze user prompts for specificity
• Detect missing required parameters
• Suggest improvements for better outputs
• Provide examples of improved prompts

When to Disable:

• Strictly minimal responses needed
• Integration with systems that can’t handle JSON
• Lower latency required (small overhead when enabled)

DEVOPS_OS_SUGGESTION_CONFIDENCE_THRESHOLD#

Type: String
Default: medium
Valid Values: high, medium, low

Minimum confidence level for suggestions to display.

Confidence Levels:

• high - Only critical gaps
• medium - Important gaps (balanced, recommended default)
• low - All suggestions including minor improvements

Example:

# Show only high-confidence suggestions
export DEVOPS_OS_SUGGESTION_CONFIDENCE_THRESHOLD=high

# Show medium and high (default, balanced)
export DEVOPS_OS_SUGGESTION_CONFIDENCE_THRESHOLD=medium

# Show all suggestions
export DEVOPS_OS_SUGGESTION_CONFIDENCE_THRESHOLD=low

Selection Guide:

• high - For production, reduce noise
• medium - Standard use, good balance
• low - Learning/exploration mode

DEVOPS_OS_MAX_SUGGESTIONS_PER_RESPONSE#

Type: Integer
Default: 2
Valid Range: 1-10

Maximum number of suggestions to include per response.

Example:

# Default: 2 suggestions
export DEVOPS_OS_MAX_SUGGESTIONS_PER_RESPONSE=2

# Show only 1 suggestion
export DEVOPS_OS_MAX_SUGGESTIONS_PER_RESPONSE=1

# Show up to 3 suggestions
export DEVOPS_OS_MAX_SUGGESTIONS_PER_RESPONSE=3

Balancing:

• 1 - Minimal, show most critical suggestion only
• 2 - Balanced (default)
• 3+ - Comprehensive guidance

DEVOPS_OS_INCLUDE_EXAMPLES_IN_SUGGESTIONS#

Type: Boolean
Default: true
Valid Values: true, false, 1, 0, yes, no

Include example improved prompts in suggestions.

Example:

# Include examples (default)
export DEVOPS_OS_INCLUDE_EXAMPLES_IN_SUGGESTIONS=true

# Exclude examples (shorter responses)
export DEVOPS_OS_INCLUDE_EXAMPLES_IN_SUGGESTIONS=false

When to Disable:

• Minimal response size needed
• Using with systems sensitive to response length
• Faster parsing preferred

When to Enable:

• Users learning how to improve prompts
• Interactive chatbot scenario
• Educational use

Configuration Examples#

Local Development (Claude Desktop)#

# Minimal configuration
export DEVOPS_OS_TRANSPORT=stdio
export DEVOPS_OS_PROFILE=local
export DEVOPS_OS_LOG_LEVEL=INFO

python -m mcp_server.server

Docker Local (HTTP, no auth)#

export DEVOPS_OS_TRANSPORT=streamable-http
export DEVOPS_OS_PROFILE=local
export DEVOPS_OS_HOST=0.0.0.0
export DEVOPS_OS_PORT=8000
export DEVOPS_OS_LOG_LEVEL=INFO

python -m mcp_server.server

Remote Production (ChatGPT, JWT auth)#

export DEVOPS_OS_TRANSPORT=streamable-http
export DEVOPS_OS_PROFILE=remote
export DEVOPS_OS_HOST=0.0.0.0
export DEVOPS_OS_PORT=8000
export DEVOPS_OS_JWT_ISSUER=https://auth.example.com/
export DEVOPS_OS_JWT_AUDIENCE=devops-os-service
export DEVOPS_OS_LOG_LEVEL=WARNING
export DEVOPS_OS_EXECUTION_TIMEOUT=60

python -m mcp_server.server

Performance Tuned (Large configurations)#

export DEVOPS_OS_TRANSPORT=streamable-http
export DEVOPS_OS_REQUEST_SIZE_MB=100
export DEVOPS_OS_RESPONSE_SIZE_MB=200
export DEVOPS_OS_EXECUTION_TIMEOUT=120
export DEVOPS_OS_LOG_LEVEL=INFO

python -m mcp_server.server

Debug Mode (Troubleshooting)#

export DEVOPS_OS_LOG_LEVEL=DEBUG
export DEVOPS_OS_EXECUTION_TIMEOUT=120
export DEVOPS_OS_ENABLE_SUGGESTIONS=true
export DEVOPS_OS_SUGGESTION_CONFIDENCE_THRESHOLD=low
export DEVOPS_OS_MAX_SUGGESTIONS_PER_RESPONSE=5

python -m mcp_server.server

Docker Compose Example#

version: '3.8'

services:
  devops-os:
    image: devops-os-mcp:latest
    environment:
      DEVOPS_OS_TRANSPORT: streamable-http
      DEVOPS_OS_PROFILE: remote
      DEVOPS_OS_HOST: 0.0.0.0
      DEVOPS_OS_PORT: 8000
      DEVOPS_OS_JWT_ISSUER: https://auth.example.com/
      DEVOPS_OS_JWT_AUDIENCE: devops-os-service
      DEVOPS_OS_LOG_LEVEL: INFO
      DEVOPS_OS_REQUEST_SIZE_MB: 50
      DEVOPS_OS_RESPONSE_SIZE_MB: 100
      DEVOPS_OS_EXECUTION_TIMEOUT: 60
    ports:
      - "8000:8000"
    restart: unless-stopped

Environment Variable Loading#

DevOps-OS MCP loads variables in this priority order:

1. Explicitly set environment variables (highest priority)
2. .env file (if present in working directory)
3. Built-in defaults (lowest priority)

Using a .env File#

Create .env file:

DEVOPS_OS_TRANSPORT=streamable-http
DEVOPS_OS_PROFILE=remote
DEVOPS_OS_JWT_ISSUER=https://auth.example.com/
DEVOPS_OS_JWT_AUDIENCE=devops-os-service

Load and run:

set -a
source .env
set +a
python -m mcp_server.server

Validation & Constraints#

Troubleshooting#

“Configuration error: Port must be 1-65535”#

• Check DEVOPS_OS_PORT is a valid port number
• Ensure no letters or special characters

“Invalid profile: remote” with missing JWT#

• When using DEVOPS_OS_PROFILE=remote, both DEVOPS_OS_JWT_ISSUER and DEVOPS_OS_JWT_AUDIENCE are required
• Set both variables before starting server

“Tool execution timeout” errors#

• Increase DEVOPS_OS_EXECUTION_TIMEOUT
• Check server system resources
• Verify network latency

Request/Response size errors#

• Increase DEVOPS_OS_REQUEST_SIZE_MB for large inputs
• Increase DEVOPS_OS_RESPONSE_SIZE_MB for large outputs
• Check server available memory

“Bind address already in use”#

• Change DEVOPS_OS_PORT to unused port
• Or stop other process using same port
• Check with: lsof -i :8000

See Also#

• MCP Setup & Configuration Guide
• Prompt Improvement Suggestions