Development MCP Servers

Free capability resolver for AI agents. Discovers and compares tools, APIs, MCP servers, and x402 machine services through a no-auth remote MCP endpoint.

Queay exposes a remote MCP server so a coding agent - Claude Code, Claude Desktop, GitHub Copilot, Kilo Code, and others - can list your test scripts, trigger a run, and check its status without leaving your editor. All execution still runs on Queay's own infrastructure; the agent only triggers and reads back results, the same as CI/CD

Cuts MCP tool-definition token overhead with lossless TOON encoding, so agents can load many tools without burning context. Zero-dependency Python CLI.

An AI agent that drives VS Code's real debugger — breakpoints, stepping, variable inspection — and an MCP server so external agents (Claude Code, Cursor) can debug through your editor too.

Test case management system TuskR

Official Featureflip server: create, toggle and target feature flags, see one flag across every environment, and find stale flags ready for cleanup.

MCP server for Redmine: lets AI agents (Claude Code, Cursor, Copilot) search issues, projects, wiki, attachments and time entries, run release and blocker analytics, and optionally create issues, notes and time entries

Read-only docs for the ADITUS event-technology APIs (ticketing, access, BI) and Shop Micro Frontend.

Three-layer firewall for MCP: inspects tool calls, content and tool results for prompt injection, credential leaks and data egress before they reach the model. Runs locally via npx.

Discover and load existing local Agent Skills through MCP. Preserve origins, resolve name collisions, and check explicit SKILL.md files in CI.

Gives coding agents a disposable Android emulator: start a session from an APK, demo app or GitHub pull request, snapshot the UI tree, tap, type, screenshot, read logcat, and share a browser watch link with a person. Streamable HTTP at https://api.ramus.dev/api/mcp with a Bearer API key; CLI on npm as ramus-cli.

File-native collaboration and governance protocol for multi-agent systems. TASK, REPORT, ISSUE and REVIEW coordinate assignments, parallel delivery and evidence-based review. No coordination database or message queue required.

Measure cost per successful task, latency and quality from your own agent traces. Three free analysis tools via MCP, with an ALPNAI key.

Can I pay this address? First 5 free. Then $0.02 USDC. No inbox.

Official MCP server of Melbis Shop, a self-hosted e-commerce platform: an AI agent works in the store as an employee — files, database, catalogue, orders — under a staff login with per-command permissions.

In-process zero-trust binary admission gate, prompt injection defense, and autonomous thermal governor for MCP agents.

Pre-1.0 governed knowledge for coding agents: markdown in git, lifecycle status, in-force filter.

MCP server for writing and text editing tools.

MCP server for working with Makefile targets.

MCP server for mutation testing with mutmut.

Wrenda exposes a per-domain MCP server over JSON-RPC so AI agents can query site content programmatically. It is also an edge reverse proxy for AI search: GPTBot/ClaudeBot/Perplexity get enriched semantic markdown; Googlebot gets pre-rendered HTML; humans get the original page. Setup is a DNS change (~15 minutes). 14-day free trial, no credit card.

Run 180+ AI models, image, video and speech generation, web search, batch jobs, GPU Cloud and hosted agents as tools. OAuth sign-in with your EmpirioLabs account.

MCP server for managing Hugo frontmatter.

180 tools that give an agent exact answers instead of guesses: IBAN and VAT validation, cron, regex, JSON, dates, Dutch test data (BSN, IBAN, BRP), PDF and image renders, agent memory and saved workflows. Remote server, 150 utilities free without a key.

AI customer support system that turns user feedback into Codex and Multica workflows, helping SaaS teams improve every week.

Remote Linux/macOS host administration — run shell commands, manage files, and check system metrics through an Auth0-secured MCP server.

Local static inspection of MCP and AI-agent tool manifests before attachment, reporting findings and missing evidence without executing proposed tools; results do not establish runtime safety.

ABAP development from Claude and other MCP hosts: search objects, read and write source, transports, activation, ABAP Unit, ATC, short dumps, debugger and SQL queries over ADT. 173 tools, multi-system (S/4HANA Cloud and on-prem), with server-side guard rails (read-only, allowed packages, denied tables) and audit log.

Zero-false-positive tripwires for AI agents: 8 inert decoy MCP tools (fake transfer, prod-secrets reader, sudo shell) plus canary tokens planted in honeypot files. Any touch is a compromise signal — prompt-injection detection with full attack context. MIT, works with Claude Code / Cursor / any MCP client.

Prices your Claude Code and Codex sessions from the transcripts on disk, shows what each agent cost over any period, and tells you in plain English what would have been cheaper. CLI, Browser UI and MCP server.