"MCP servers for reducing hallucinations while coding" matching MCP servers:

Voice interface for Claude Code: you talk, the agent listens, codes, and talks back while it works. Live speech-to-text with turn-taking, Grok/xAI voices with per-subagent personas, and a real-time HUD dashboard.

Enables continuous voice conversation with AI coding assistants by locally transcribing speech with Whisper and delivering utterances as text prompts.

Enables AI video dubbing by letting users submit video URLs for dubbing into Russian, English, or Spanish while preserving original speakers' voices through voice cloning, plus job status and cost estimation tools.

Enables AI agents to transcribe audio, detect ASR hallucinations deterministically, repair flagged segments with guarded LLM calls, score transcript quality, and estimate pipeline costs over stdio.

Enables AI assistants to transcribe audio and video from URLs or local files with high accuracy, speaker diarization, 119 languages, and word-level timestamps, while also supporting transcription management and caption export in SRT, WebVTT, or plain text.

Turns text into local audio files (MP3) via OpenRouter speech models, enabling AI coding agents to generate voiceover narrations for video production.

Enables voice input for AI coding assistants by recording audio, transcribing it locally with Whisper, and optionally typing the text at the cursor position. Supports multiple languages and works with MCP-compatible tools like OpenCode, Claude Code, and Cursor.

Connects to the TypeWhisper macOS app to let coding agents transcribe local files, inspect model status, search history, and manage dictionary terms and corrections.

Lets cloud-based coding agents send progress updates and read messages that a macOS voice app speaks aloud and accepts spoken replies for, with no installation on the agent side. Agents can report status, fetch instructions dictated by voice, and acknowledge them through a hosted endpoint.

A Model Context Protocol server that provides voice-AI session management and usage tracking tools for SpekoAI. It enables creating, retrieving, and ending voice sessions while monitoring usage through MCP clients like Claude Desktop.

Enables coding agents to participate in voice calls, speaking replies while continuing their work and accepting spoken input or interruptions within the same conversation. It delivers voice input into Claude Code, Codex, Cursor, or custom HTTP endpoint workflows.

Enables coding agents to build, test, and deploy production voice agents with real phone numbers directly from the editor, including live call monitoring and sandboxed development.

Enables AI coding agents and MCP clients to generate Giggy text-to-speech audio by listing voices, requesting speech generations, and polling generation status until a downloadable result URL is ready.

Read-only MCP server for Telegram that enables reading messages and transcribing voice, audio, and video notes via the Telegram API for use with Codex and Claude Code.

Enables small online store owners to run a Shopify or WooCommerce shop by voice through Alexa+, answering questions about sales and stock while executing confirmation-gated actions like shipping orders or drafting supplier messages. Destructive or large-impact writes require a spoken code and are recorded in an audit log.

Local-first voice I/O for MCP clients, enabling hands-free dictation, spoken responses, and voice-based Q&A with coding agents.

Enables a coding agent to see and hear the host machine via webcam stills, microphone speech-to-text, and optional text-to-speech replies, all processed locally.

Enables hands-free voice collaboration with a coding agent: you speak requirements aloud and hear only short, plain-language progress and completion summaries while the agent edits code, runs tests, and fixes bugs in the background. It also handles continuous speech recognition with automatic turn detection, barge-in interruption, and spoken approval prompts before risky operations.

Enables an Alexa+ assistant to name scam warning signs drawn from FTC and FBI alerts, verify a story by texting a family member's pre-saved number, and surface official report links and guidance. It exposes tools and MCP Apps views for checking a suspicious call, while never approving payments, contacting the caller, or collecting sensitive numbers.

An MCP server that watches for screen recordings, transcribes spoken feedback locally, and attaches each comment to the relevant frame, enabling coding agents to see and fix the issues you mention.