Skip to main content
Glama
Gorav22

TerminusAI

by Gorav22
OverviewSchemaSourceRelated ServersDiscussionsAdmin
TypeScript

execute_command

Execute terminal commands locally or on remote hosts via SSH with session persistence and environment variable support for efficient command workflows.

Instructions

Execute commands on remote hosts or locally (This tool can be used for both remote hosts and the current machine)

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault
hostNoHost to connect to (optional, if not provided the command will be executed locally)
usernameNoUsername for SSH connection (required when host is specified)
sessionNoSession name, defaults to 'default'. The same session name will reuse the same terminal environment for 20 minutes, which is useful for operations requiring specific environments like conda.default
commandYesCommand to execute. Before running commands, it's best to determine the system type (Mac, Linux, etc.)
envNoEnvironment variables

Implementation Reference

  • src/executor.ts:128-320 (handler)
    Main handler that executes the command locally or on remote host via SSH, managing persistent sessions, environment variables, and using either interactive shell or exec.
    async executeCommand(
  command: string,
  options: {
    host?: string;
    username?: string;
    session?: string;
    env?: Record<string, string>;
  } = {}
): Promise<{stdout: string; stderr: string}> {
  const { host, username, session = 'default', env = {} } = options;
  const sessionKey = this.getSessionKey(host, session);

  // 如果指定了host，则使用SSH执行命令
  if (host) {
    if (!username) {
      throw new Error('Username is required when using SSH');
    }
    
    let sessionData = this.sessions.get(sessionKey);
    
    let needNewConnection = false;
    if (!sessionData || sessionData.host !== host) {
      needNewConnection = true;
    } else if (sessionData.client) {
      if (sessionData.client.listenerCount('ready') === 0 && sessionData.client.listenerCount('data') === 0) {
        log.info(`Session ${sessionKey} disconnected, reconnecting`);
        needNewConnection = true;
      }
    } else {
      needNewConnection = true;
    }
    
    if (needNewConnection) {
      log.info(`Creating new connection for command execution: ${sessionKey}`);
      await this.connect(host, username, session);
      sessionData = this.sessions.get(sessionKey);
    } else {
      log.info(`Reusing existing session for command execution: ${sessionKey}`);
    }
    
    if (!sessionData || !sessionData.client) {
      throw new Error(`host ${host} err`);
    }
    
    this.resetTimeout(sessionKey);

    if (sessionData.shellReady && sessionData.shell) {
      log.info(`Executing command using interactive shell: ${command}`);
      return new Promise((resolve, reject) => {
        let stdout = "";
        let stderr = "";
        let commandFinished = false;
        const uniqueMarker = `CMD_END_${Date.now()}_${Math.random().toString(36).substring(2, 15)}`;
        
        const envSetup = Object.entries(env)
          .map(([key, value]) => `export ${key}="${String(value).replace(/"/g, '\\"')}"`)
          .join(' && ');
        
        const fullCommand = envSetup ? `${envSetup} && ${command}` : command;
        
        const dataHandler = (data: Buffer) => {
          const str = data.toString();
          log.debug(`Shell数据: ${str}`);
          
          if (str.includes(uniqueMarker)) {
            commandFinished = true;
            
            const lines = stdout.split('\n');
            let commandOutput = '';
            let foundCommand = false;
            
            for (const line of lines) {
              if (foundCommand) {
                if (line.includes(uniqueMarker)) {
                  break;
                }
                commandOutput += line + '\n';
              } else if (line.includes(fullCommand)) {
                foundCommand = true;
              }
            }
            
            resolve({ stdout: commandOutput.trim(), stderr });
            
            sessionData.shell.removeListener('data', dataHandler);
            clearTimeout(timeout);
          } else if (!commandFinished) {
            stdout += str;
          }
        };
        
        const errorHandler = (err: Error) => {
          stderr += err.message;
          reject(err);
          sessionData.shell.removeListener('data', dataHandler);
          sessionData.shell.removeListener('error', errorHandler);
        };
        
        sessionData.shell.on('data', dataHandler);
        sessionData.shell.on('error', errorHandler);
        
        sessionData.shell.write(`echo "Starting command execution: ${fullCommand}"\n`);
        sessionData.shell.write(`${fullCommand}\n`);
        sessionData.shell.write(`echo "${uniqueMarker}"\n`);
        
        const timeout = setTimeout(() => {
          if (!commandFinished) {
            stderr += "Command execution timed out";
            resolve({ stdout, stderr });
            sessionData.shell.removeListener('data', dataHandler);
            sessionData.shell.removeListener('error', errorHandler);
          }
        }, 30000); // 30 seconds
      });
    } else {
      log.info(`Executing command using exec: ${command}`);
      return new Promise((resolve, reject) => {
        const envSetup = Object.entries(env)
          .map(([key, value]) => `export ${key}="${String(value).replace(/"/g, '\\"')}"`)
          .join(' && ');
        
        const fullCommand = envSetup ? `${envSetup} && ${command}` : command;
        
        sessionData?.client?.exec(`/bin/bash --login -c "${fullCommand.replace(/"/g, '\\"')}"`, (err, stream) => {
          if (err) {
            reject(err);
            return;
          }

          let stdout = "";
          let stderr = '';

          stream
            .on("data", (data: Buffer) => {
              this.resetTimeout(sessionKey);
              stdout += data.toString();
            })
            .stderr.on('data', (data: Buffer) => {
              stderr += data.toString();
            })
            .on('close', () => {
              resolve({ stdout, stderr });
            })
            .on('error', (err) => {
              reject(err);
            });
        });
      });
    }
  } 
  else {
    log.info(`Executing command using local session: ${sessionKey}`);
    
    let sessionData = this.sessions.get(sessionKey);
    let sessionEnv = {};
    
    if (!sessionData) {
      sessionData = {
        client: null,
        connection: null,
        timeout: null,
        host: undefined,
        env: { ...env } 
      };
      this.sessions.set(sessionKey, sessionData);
      log.info(`Creating new local session: ${sessionKey}`);
      sessionEnv = env;
    } else {
      log.info(`Reusing existing local session: ${sessionKey}`);
      if (!sessionData.env) {
        sessionData.env = {};
      }
      sessionData.env = { ...sessionData.env, ...env };
      sessionEnv = sessionData.env;
      this.sessions.set(sessionKey, sessionData);
    }
    
    this.resetTimeout(sessionKey);
    
    return new Promise((resolve, reject) => {
      const envVars = { ...process.env, ...sessionEnv };
      
      log.info(`Executing local command: ${command}`);
      exec(command, { env: envVars }, (error, stdout, stderr) => {
        if (error && error.code !== 0) {
          resolve({ stdout, stderr: stderr || error.message });
        } else {
          resolve({ stdout, stderr });
        }
      });
    });
  }
}
  • src/index.ts:49-75 (schema)
    Input schema definition for the execute_command tool, specifying parameters like host, username, session, command, and env.
    inputSchema: {
  type: "object",
  properties: {
    host: {
      type: "string",
      description: "Host to connect to (optional, if not provided the command will be executed locally)"
    },
    username: {
      type: "string",
      description: "Username for SSH connection (required when host is specified)"
    },
    session: {
      type: "string",
      description: "Session name, defaults to 'default'. The same session name will reuse the same terminal environment for 20 minutes, which is useful for operations requiring specific environments like conda.",
      default: "default"
    },
    command: {
      type: "string",
      description: "Command to execute. Before running commands, it's best to determine the system type (Mac, Linux, etc.)"
    },
    env: {
      type: "object",
      description: "Environment variables",
      default: {}
    }
  },
  required: ["command"]
  • src/index.ts:43-80 (registration)
    Registers the execute_command tool in the MCP server's ListToolsRequest handler by providing its name, description, and schema.
    server.setRequestHandler(ListToolsRequestSchema, async () => {
  return {
    tools: [
      {
        name: "execute_command",
        description: "Execute commands on remote hosts or locally (This tool can be used for both remote hosts and the current machine)",
        inputSchema: {
          type: "object",
          properties: {
            host: {
              type: "string",
              description: "Host to connect to (optional, if not provided the command will be executed locally)"
            },
            username: {
              type: "string",
              description: "Username for SSH connection (required when host is specified)"
            },
            session: {
              type: "string",
              description: "Session name, defaults to 'default'. The same session name will reuse the same terminal environment for 20 minutes, which is useful for operations requiring specific environments like conda.",
              default: "default"
            },
            command: {
              type: "string",
              description: "Command to execute. Before running commands, it's best to determine the system type (Mac, Linux, etc.)"
            },
            env: {
              type: "object",
              description: "Environment variables",
              default: {}
            }
          },
          required: ["command"]
        }
      }
    ]
  };
});
  • src/index.ts:82-133 (handler)
    MCP CallTool request handler specific to execute_command, validates params, invokes CommandExecutor, and formats response.
    server.setRequestHandler(CallToolRequestSchema, async (request) => {
  try {
    if (request.params.name !== "execute_command") {
      throw new McpError(ErrorCode.MethodNotFound, "Unknown tool");
    }
    
    const host = request.params.arguments?.host ? String(request.params.arguments.host) : undefined;
    const username = request.params.arguments?.username ? String(request.params.arguments.username) : undefined;
    const session = String(request.params.arguments?.session || "default");
    const command = String(request.params.arguments?.command);
    if (!command) {
      throw new McpError(ErrorCode.InvalidParams, "Command is required");
    }
    const env = request.params.arguments?.env || {};

    if (host && !username) {
      throw new McpError(ErrorCode.InvalidParams, "Username is required when host is specified");
    }

    try {
      const result = await commandExecutor.executeCommand(command, {
        host,
        username,
        session,
        env: env as Record<string, string>
      });
      
      return {
        content: [{
          type: "text",
          text: `Command Output:\nstdout: ${result.stdout}\nstderr: ${result.stderr}`
        }]
      };
    } catch (error) {
      if (error instanceof Error && error.message.includes('SSH')) {
        throw new McpError(
          ErrorCode.InternalError,
          `SSH connection error: ${error.message}. Please ensure SSH key-based authentication is set up.`
        );
      }
      throw error;
    }
  } catch (error) {
    if (error instanceof McpError) {
      throw error;
    }
    throw new McpError(
      ErrorCode.InternalError,
      error instanceof Error ? error.message : String(error)
    );
  }
});

Schema Changelog

Changes observed during successful MCP inspections.

  1. First observed

TDQS

B3.1/5.0
Behavior2/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

With no annotations provided, the description carries the full burden of behavioral disclosure. It mentions that commands can be executed remotely or locally and hints at session reuse for 20 minutes, but it lacks critical details such as security implications, error handling, output format, or potential side effects. This leaves significant gaps for a tool that executes arbitrary commands.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is a single, efficient sentence that front-loads the core functionality. It avoids unnecessary details and stays focused on the tool's scope, though it could be slightly more structured by explicitly separating remote and local use cases.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness2/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given the complexity of executing commands (which can have security and system impacts), the lack of annotations, and no output schema, the description is insufficient. It fails to address critical aspects like return values, error conditions, or safety warnings, making it incomplete for informed tool usage.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

The input schema has 100% description coverage, so the baseline is 3. The description does not add any meaningful parameter semantics beyond what is already documented in the schema, such as explaining the 'command' parameter's system dependencies or the 'session' parameter's environmental implications in more depth.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the tool's purpose as executing commands on remote hosts or locally, specifying both the verb ('execute') and the resource ('commands'). It distinguishes between remote and local execution contexts. However, without sibling tools, it cannot demonstrate differentiation from alternatives, preventing a perfect score.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines3/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description implies usage by mentioning both remote and local contexts, but it does not provide explicit guidance on when to choose one over the other or any prerequisites. For example, it notes host is optional for local execution but does not clarify when remote execution is preferable or what conditions might affect it.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Deploy Server

Other Tools