MCP Tools Reference

Station provides 41 MCP tools for complete agent management, execution, evaluation, and infrastructure control. All tools are available through your AI assistant (Claude, Cursor, OpenCode, etc.).

Tool Categories

Category	Tools	Purpose
Agent Management	11	Create, configure, and manage agents
Agent Execution	4	Run agents and inspect results
Evaluation & Testing	7	Test and benchmark agents
Reports & Analytics	4	Performance reporting
Environment Management	3	Manage environments
MCP Server Management	5	Configure MCP servers
Tool Discovery	2	Find available tools
Scheduling	3	Automate agent runs
Bundles	1	Package agents
Faker System	1	Mock data generation

Agent Management

`create_agent`

Create a new agent with prompts, tools, and configuration.

"Create a cost-analyzer agent that uses AWS tools to analyze spending"

Parameters:

name (required) - Agent name
description (required) - What the agent does
prompt (required) - System prompt
environment_id - Target environment
max_steps - Maximum execution steps (default: 5)
tool_names - List of tools to assign
sandbox - Sandbox configuration (python/node/bash)
output_schema - JSON schema for structured output

`update_agent`

Modify an existing agent’s configuration.

"Update cost-analyzer to use max_steps 10"

`update_agent_prompt`

Update only the agent’s system prompt.

"Update the prompt for incident-coordinator to be more concise"

`delete_agent`

Remove an agent permanently.

"Delete the old-analyzer agent"

`list_agents`

List all agents with optional filters.

"List all agents in the production environment"

Parameters:

environment_id - Filter by environment
enabled_only - Show only enabled agents
limit - Max results (default: 50)
offset - Pagination offset

`get_agent_details`

Get complete configuration for a specific agent.

"Show me details for the incident-coordinator agent"

`get_agent_schema`

Get the input schema for an agent’s dotprompt template.

"What variables does cost-analyzer accept?"

`add_tool`

Assign an MCP tool to an agent.

"Add the __aws_cost_explorer tool to cost-analyzer"

`remove_tool`

Remove a tool from an agent.

"Remove __deprecated_tool from my-agent"

`add_agent_as_tool`

Create multi-agent hierarchies by adding an agent as a callable tool.

"Add logs-investigator as a tool for incident-coordinator"

`remove_agent_as_tool`

Break an agent hierarchy link.

"Remove logs-investigator from incident-coordinator's tools"

Agent Execution

`call_agent`

Execute an agent with a specific task.

"Run cost-analyzer with task: Analyze last week's AWS spending"

Parameters:

agent_id (required) - Agent to execute
task (required) - Task/prompt to run
variables - Input variables for dotprompt
async - Run asynchronously (default: false)
store_run - Save to history (default: true)
timeout - Timeout in seconds (default: 300)

`list_runs`

List agent execution history.

"Show me the last 20 runs for cost-analyzer"

Parameters:

agent_id - Filter by agent
status - Filter by status (success/error/running)
limit - Max results (default: 50)

`inspect_run`

Get detailed information about a specific run.

"Show me details for run 123"

Parameters:

run_id (required) - Run to inspect
verbose - Include full tool calls (default: true)

`list_runs_by_model`

Filter runs by AI model.

"Show me all runs using gpt-4o-mini"

Evaluation & Testing

`generate_and_test_agent`

Generate test scenarios and run comprehensive agent testing.

"Generate 50 test scenarios for incident-coordinator and run them"

Parameters:

agent_id (required) - Agent to test
scenario_count - Number of scenarios (default: 100)
max_concurrent - Parallel executions (default: 10)
variation_strategy - comprehensive/edge_cases/common

`batch_execute_agents`

Run multiple agents concurrently for testing.

"Run cost-analyzer and security-scanner in parallel with 5 iterations each"

Parameters:

tasks - JSON array of execution tasks
iterations - Times to run each (default: 1, max: 100)
max_concurrent - Parallel limit (default: 5, max: 20)

`evaluate_benchmark`

Run LLM-as-judge evaluation on a completed run.

"Evaluate run 123 using the benchmark system"

`evaluate_dataset`

Evaluate an entire dataset of agent runs.

"Evaluate the dataset at /path/to/dataset"

`export_dataset`

Export runs for external analysis.

"Export the last 100 runs to Genkit format"

Parameters:

filter_agent_id - Filter by agent
filter_model - Filter by model
limit - Max runs (default: 100)
output_dir - Output directory (default: ./evals/)

`list_benchmark_results`

List previous evaluation results.

"Show me benchmark results for run 123"

`get_benchmark_status`

Check status of an evaluation task.

"What's the status of benchmark task abc123?"

Reports & Analytics

`create_report`

Create a new performance report definition.

"Create a weekly report for the SRE team measuring incident response time"

Parameters:

name (required) - Report name
environment_id (required) - Environment to evaluate
team_criteria (required) - JSON defining success metrics
agent_criteria - Per-agent metrics
filter_model - Filter by AI model

`generate_report`

Run benchmarks and generate a report.

"Generate the SRE team performance report"

`list_reports`

List all report definitions.

"Show me all reports for production"

`get_report`

Get details of a specific report.

"Show me the latest SRE report results"

Environment Management

`create_environment`

Create a new environment.

"Create a staging environment"

`delete_environment`

Delete an environment and all its data.

"Delete the old-staging environment"

`list_environments`

List all environments.

"What environments do I have?"

MCP Server Management

`add_mcp_server_to_environment`

Add an MCP server configuration.

"Add a Datadog MCP server to production"

Parameters:

environment_name (required)
server_name (required)
command (required) - Server command
args - Command arguments
env - Environment variables
description - Server description

`update_mcp_server_in_environment`

Update an existing MCP server.

"Update the datadog server to use a new API key"

`delete_mcp_server_from_environment`

Remove an MCP server.

"Remove the deprecated-server from production"

`list_mcp_servers_for_environment`

List servers in an environment.

"What MCP servers are in production?"

`list_mcp_configs`

List all MCP configurations.

"Show me all MCP configs across environments"

Tool Discovery

`discover_tools`

Discover tools from MCP server configurations.

"Discover what tools are available from the AWS server"

Parameters:

environment_id - Filter by environment
config_id - Specific MCP config

`list_tools`

List available tools with search.

"List all tools that mention 'cost'"

Parameters:

environment_id - Filter by environment
search - Search term
limit - Max results (default: 50)

Scheduling

`set_schedule`

Schedule an agent to run automatically.

"Schedule cost-analyzer to run daily at 9 AM"

Parameters:

agent_id (required)
cron_schedule (required) - 6-field cron expression
schedule_variables - Variables for scheduled runs
enabled - Enable/disable (default: true)

`get_schedule`

View an agent’s schedule.

"What's the schedule for cost-analyzer?"

`remove_schedule`

Remove an agent’s schedule.

"Remove the schedule from cost-analyzer"

Bundles

`create_bundle_from_environment`

Package an environment as a distributable bundle.

"Create a bundle from the production environment"

Parameters:

environmentName (required)
outputPath - Output file path

Faker System

`faker_create_standalone`

Create an AI-powered mock data server.

"Create a prometheus faker that generates Kubernetes metrics"

Parameters:

environment_name (required)
faker_name (required)
description (required)
goal (required) - AI instruction for data generation
tools - JSON array of tool definitions
auto_sync - Sync to environment (default: true)
persist - Persist to template.json (default: true)

Usage Examples

Create an Incident Response Team

You: Create an incident response team with a coordinator and three specialists

Claude: I'll create the team for you.

[Using create_agent with name="incident-coordinator"...]
✅ Created incident_coordinator

[Using create_agent with name="logs-investigator"...]
✅ Created logs_investigator

[Using add_agent_as_tool...]
✅ Added logs_investigator as tool for incident_coordinator

[Continues for all specialists...]

Your incident response team is ready!

Run a Cost Analysis

You: Analyze yesterday's AWS costs

Claude: I'll run the cost analyzer.

[Using call_agent with agent_id="21", task="Analyze yesterday's costs"...]

Results:
- Total spend: $1,247.32
- Top service: EC2 ($580)
- 15% increase from last week
- Anomaly detected: Lambda costs up 40%

Generate Performance Report

You: How is my SRE team performing?

Claude: [Using generate_report with report_id="sre-team"...]

Team Performance: 7.5/10
- Incident response: 8.5/10
- Tool utilization: 8.0/10
- Root cause accuracy: 7.5/10
- Communication: 6.5/10

Tool Naming Convention

Station tools appear with prefix based on your MCP client
In OpenCode: opencode-station_create_agent
In Claude Desktop: station-create_agent
Tool names from MCP servers use __ prefix: __aws_cost_explorer

Next Steps

Agent Configuration - Detailed agent setup
Scheduling - Automate agent runs
Observability - Monitor agent execution