SentryFrogg MCP Server

SentryFrogg MCP Server

Provides a governed command plane for MCP agents to manage PostgreSQL databases, SSH sessions, and HTTP APIs with encrypted credential storage and prescriptive workflows.

Category
Visit Server

README

SentryFrogg MCP Server v4.2.0

Document Profile

Field Value
Product SentryFrogg MCP Server
Version 4.2.0
Runtime Node.js ≥ 16
Interfaces Model Context Protocol (PostgreSQL · SSH · HTTP)
License MIT

Executive Synopsis

SentryFrogg MCP Server supplies a governed command plane for MCP-compatible agents that operate data platforms, remote shells and web services. The implementation prioritises deterministic behaviour, encrypted credential storage and prescriptive workflows so that autonomous agents can execute infrastructure tasks with enterprise auditability.

Functional Surface

Capability Detail
PostgreSQL control Parameterised SQL, catalog discovery, CRUD helpers, mutual TLS, connection profiling.
SSH execution Serial command dispatch per profile, password/key authentication, reachability diagnostics.
HTTP access REST verbs with JSON payloads, health checks, programmable headers and bearer tokens.
Secret vault AES-256 encrypted profiles persisted in profiles.json with managed key material.
Telemetry Tool-level statistics via getStats() for integration with monitoring stacks.

System Components

Component Scope
sentryfrogg_server.cjs MCP entry point, tool catalogue, lifecycle supervision.
src/bootstrap/ServiceBootstrap.cjs Service registration, dependency wiring, resource cleanup.
src/managers/PostgreSQLManager.cjs SQL execution, profile validation, TLS configuration, pool management.
src/managers/SSHManager.cjs SSH sessions, sequential execution enforcement, profile hygiene.
src/managers/APIManager.cjs HTTP invocation, header synthesis, response shaping.
src/services/ProfileService.cjs Encrypted persistence and retrieval of profile objects.
src/services/Security.cjs Key lifecycle, cryptographic primitives, payload size guards.
src/services/Validation.cjs Canonical validation for incoming payloads.

MCP Tools

mcp_psql_manager

Attribute Specification
Required flow setup_profile → downstream action using same profile_name.
Actions setup_profile, list_profiles, quick_query, show_tables, describe_table, sample_data, insert_data, update_data, delete_data, database_info.
Credentials Either discrete fields (host, port, username, password, database) or connection_url.
TLS options ssl_mode, ssl_ca, ssl_cert, ssl_key, ssl_passphrase, ssl_servername, ssl_reject_unauthorized; omitted values reuse stored secrets.
Query rules quick_query injects LIMIT 100 when absent; bind variables supplied via params array ($1, $2, ...).
Response format JSON with success, rows, rowCount, fields, command; failures raise MCP internal errors.
Rejection triggers Missing profile, malformed SQL, payload limits exceeded, invalid TLS configuration.

mcp_ssh_manager

Attribute Specification
Required flow setup_profile (password or PEM private_key, optional passphrase) → operational action.
Actions setup_profile, list_profiles, execute, system_info, check_host.
Execution model Commands trimmed and length-limited; pipes/redirects permitted; per-profile execution is strictly sequential.
Outputs JSON containing success, stdout, stderr, exitCode, durationMs; errors propagate as MCP internal errors.
Security posture Secrets encrypted at rest; no templating—agents must supply fully qualified commands.

mcp_api_client

Attribute Specification
Actions get, post, put, delete, patch, check_api.
Inputs url (required), data (JSON body for mutating verbs), headers (string map), auth_token (prefixed into Authorization unless already set).
Behaviour Local and private addresses allowed; HTTP status/body returned in structured JSON; transport or parsing failures emit MCP internal errors.

Profile Lifecycle

  1. Invoke setup_profile to persist credentials and TLS artefacts; secrets encrypt with AES-256 using .mcp_profiles.key (0600 permissions).
  2. Reference the same profile_name for subsequent operations; omitted sensitive fields inherit stored encrypted values.
  3. Rotate credentials by reissuing setup_profile; the latest payload supersedes previous entries.
  4. Audit existing profiles via list_profiles; responses never disclose secrets.
  5. Retire unused profiles by editing profiles.json under change control.

TLS Configuration Guidance

  • Prefer embedding sslmode directives in connection_url; explicit payload fields override URL parameters.
  • Keep ssl_reject_unauthorized at true unless communicating with trusted self-signed endpoints.
  • Provide ssl_servername whenever certificate CN/SAN mismatches the host.
  • Supply PEM blocks as single-line strings using \n escape sequences; leading/trailing spaces are disallowed.
  • ssl_passphrase must be non-empty if provided; omit otherwise.

Installation and Operations

Task Command
Clone and install git clone https://github.com/yourusername/sentryfrogg-mcp.git && cd sentryfrogg-mcp && npm install
Syntax check npm run check
Launch (stdio) node sentryfrogg_server.cjs
Update dependencies npm install --package-lock-only && npm audit fix --only=prod (subject to governance)
Reset profile store Remove profiles.json after confirming backups

Security & Compliance

  • Encryption key lifecycle: .mcp_profiles.key generated on first run; override via ENCRYPTION_KEY for coordinated environments.
  • Secret exposure: MCP responses never include decrypted values; rotation requires explicit setup_profile.
  • Input governance: SQL statements, SSH commands and HTTP payloads are length-limited; oversized inputs are rejected pre-execution.
  • Audit trail: stderr logging captures timestamped events per tool to support collection by SIEM platforms.
  • Dependency governance: locked versions of pg, ssh2, node-fetch, @modelcontextprotocol/sdk; monitor advisories for patch cadence.

Troubleshooting Matrix

Symptom Diagnostic Actions Remediation
PostgreSQL TLS failure Inspect ssl_mode, ssl_servername, certificate chain, Postgres logs. Update TLS materials; rerun setup_profile.
SSH command hang Validate command length, ensure non-interactive execution, check remote prompts. Adjust command or script; rerun setup_profile if credentials changed.
HTTP error response Review returned status/body, verify headers and auth_token. Correct payload; retry request.
Missing profile Execute list_profiles to confirm presence; ensure consistent profile_name. Recreate via setup_profile.

Change History Reference

Consult CHANGELOG.md for a dated record of functional and operational updates, including TLS support and renaming.

Contribution & Support

  • Submit changes through pull requests accompanied by verification evidence (npm run check).
  • Never commit .mcp_profiles.key or environment-specific secrets.
  • Use maintainer contact information in package.json for escalation or integration assistance.

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured