What is MCP? Model Context Protocol (MCP) is an open standard for connecting AI agents to external systems. Think of it as a universal adapter for your infrastructure.MCP servers handle authentication and permissions. Agents never see your credentials directly.
Why MCP? Approach Security Maintenance Ecosystem Direct API Full access Custom code Limited MCP Protocol Fine-grained Standard interface 100+ servers
Architecture Agent → Station MCP Client → MCP Server → External Service ↓ Validates permissions
Key Benefits:
Agents only see results, never credentials
Grant specific operations (read vs write)
Same servers work with Claude Desktop, Cursor, and Station
Configuring MCP Servers Add servers to your environment’s template.json:
{ "mcpServers" : { "filesystem" : { "command" : "npx" , "args" : [ "-y" , "@modelcontextprotocol/server-filesystem@latest" , "/workspace" ] }, "github" : { "command" : "npx" , "args" : [ "-y" , "@modelcontextprotocol/server-github" ], "env" : { "GITHUB_PERSONAL_ACCESS_TOKEN" : "{{GITHUB_TOKEN}}" } } } }
Popular MCP Servers Filesystem { "filesystem" : { "command" : "npx" , "args" : [ "-y" , "@modelcontextprotocol/server-filesystem@latest" , "/path/to/allow" ] } }
GitHub { "github" : { "command" : "npx" , "args" : [ "-y" , "@modelcontextprotocol/server-github" ], "env" : { "GITHUB_PERSONAL_ACCESS_TOKEN" : "{{GITHUB_TOKEN}}" } } }
PostgreSQL { "postgres" : { "command" : "npx" , "args" : [ "-y" , "@modelcontextprotocol/server-postgres" , "{{DATABASE_URL}}" ] } }
AWS { "aws" : { "command" : "uvx" , "args" : [ "mcp-server-aws" ], "env" : { "AWS_ACCESS_KEY_ID" : "{{AWS_ACCESS_KEY_ID}}" , "AWS_SECRET_ACCESS_KEY" : "{{AWS_SECRET_ACCESS_KEY}}" , "AWS_REGION" : "us-east-1" } } }
Using Template Variables Store secrets in variables.yml:
GITHUB_TOKEN : ghp_xxxxxxxxxxxx DATABASE_URL : postgresql://user:pass@localhost/db AWS_ACCESS_KEY_ID : AKIA... AWS_SECRET_ACCESS_KEY : secret...
Reference in template.json with {{VARIABLE_NAME}}:
{ "env" : { "GITHUB_PERSONAL_ACCESS_TOKEN" : "{{GITHUB_TOKEN}}" } }
After configuring MCP servers:
# List all available tools stn mcp tools # Sync tools (after config changes) stn env sync
Reference tools by name in your agent’s tools array:
--- tools : - "__read_file" - "__write_file" - "__search_files" - "__github_create_issue" ---
Agents can only use tools explicitly listed in their tools array.
When you connect Station as an MCP server to your AI assistant (Claude Desktop, Cursor, etc.), Station provides 41 built-in tools for managing agents:
Tool Description create_agentCreate new agents with prompts and tools update_agentModify agent configuration update_agent_promptUpdate agent system prompt delete_agentRemove an agent list_agentsList all agents (with filters) get_agent_detailsGet full agent configuration get_agent_schemaGet agent’s input schema add_toolAdd MCP tool to agent remove_toolRemove tool from agent add_agent_as_toolCreate multi-agent hierarchies remove_agent_as_toolBreak agent hierarchy links
Tool Description call_agentExecute an agent with task list_runsList agent execution history inspect_runGet detailed run information list_runs_by_modelFilter runs by AI model
Tool Description generate_and_test_agentGenerate test scenarios and run agent batch_execute_agentsRun multiple agents in parallel evaluate_benchmarkRun LLM-as-judge evaluation evaluate_datasetEvaluate entire dataset export_datasetExport runs for analysis list_benchmark_resultsList evaluation results get_benchmark_statusCheck evaluation status
Tool Description create_reportCreate team performance report generate_reportRun benchmarks and generate report list_reportsList all reports get_reportGet report details
Tool Description create_environmentCreate new environment delete_environmentDelete environment list_environmentsList all environments
Tool Description add_mcp_server_to_environmentAdd MCP server config update_mcp_server_in_environmentUpdate MCP server delete_mcp_server_from_environmentRemove MCP server list_mcp_servers_for_environmentList configured servers list_mcp_configsList all MCP configurations
Tool Description discover_toolsDiscover tools from MCP servers list_toolsList available tools
Tool Description set_scheduleSchedule agent with cron expression remove_scheduleRemove agent schedule get_scheduleGet agent schedule details
Tool Description create_bundle_from_environmentPackage environment as bundle
Tool Description faker_create_standaloneCreate AI-powered mock data server
To see all available tools in your AI assistant, just ask: “Show me all Station MCP tools available”
Next Steps
Agent Development Create agents that use these tools
Multi-Agent Teams Build teams with specialized tools
