TextToSpeechSkills Agent

Agent integrations for TextToSpeechSkills: MCP tools, an Agent Skill, and setup helpers for Codex, Claude, and Cursor.

Official links:

• Website: texttospeechskills.com
• Docs: texttospeechskills.com/docs
• Pricing: texttospeechskills.com/pricing
• API keys and MCP setup: texttospeechskills.com/app/api

Install the agent package with:

npx -y @postmastersoftware/texttospeechskills-agent setup --agent codex

That single setup command installs both pieces:

• MCP server config, so the agent can call TextToSpeechSkills tools.
• SKILL.md, so the agent knows the safe workflow for expression markup, templates, credit estimates, and generation.

Quick Start

Create a TextToSpeechSkills API key in the app, then set it in your shell:

export TTS_API_KEY=tts_live_...

Then run one of:

npx -y @postmastersoftware/texttospeechskills-agent setup --agent codex
npx -y @postmastersoftware/texttospeechskills-agent setup --agent claude
npx -y @postmastersoftware/texttospeechskills-agent setup --agent cursor

Preview first:

npx -y @postmastersoftware/texttospeechskills-agent setup --agent codex --dry-run --print-config

Run only the MCP server:

npx -y @postmastersoftware/texttospeechskills-agent mcp

The legacy binary name is also available:

npx -y --package @postmastersoftware/texttospeechskills-agent tts-skills-mcp

You can pass --api-key directly for automation, but prefer TTS_API_KEY for manual setup so the key is less likely to land in shell history.

What Gets Installed

Codex

• Skill: ~/.agents/skills/texttospeechskills/SKILL.md
• MCP: ~/.codex/config.toml
• If ~/.codex/skills already exists, the setup also writes a compatibility copy there.

Claude Code

• Skill: ~/.claude/skills/texttospeechskills/SKILL.md
• MCP: configured through claude mcp add --scope user

Cursor

• Skill: ~/.cursor/skills/texttospeechskills/SKILL.md
• MCP: ~/.cursor/mcp.json

MCP Tools

The MCP server exposes:

• tts_generate
• tts_preview
• tts_validate_expression_markup
• tts_list_tags
• tts_list_voice_templates
• tts_list_baseline_voices
• tts_suggest_voice_template
• tts_create_voice_template
• tts_estimate_usage
• tts_get_audio

The MCP server also exposes resources for expression tags, voice templates, and billing metadata:

• tts://tags
• tts://voice-templates
• tts://usage

And prompt templates:

• tts_narrate_script
• tts_make_character_voice
• tts_prepare_podcast_clip

Generation Modes

tts_generate defaults to Economy mode for queued jobs. Economy mode defaults to WAV output and currently supports single-speaker jobs.

Use Instant mode for short previews, fast turnaround, MP3 output, or two-speaker generation. tts_preview always uses Instant mode and limits text to 500 characters.

Voice Template Approval

MCP and API generation only use approved voice templates. tts_list_voice_templates returns templates that are approved for the connected workspace; draft templates stay private to the app so they can be reviewed and edited before an agent can use them.

tts_create_voice_template defaults to status: "draft". To approve a draft, open the TextToSpeechSkills app, go to Voice templates, edit the template, set Status to Approved, and save. Once approved, the template appears in tts_list_voice_templates and can be used by tts_generate or tts_preview.

Only create a template with status: "approved" when it is ready for agent/API use.

Environment

The MCP server reads:

• TTS_API_KEY or TEXTTOSPEECHSKILLS_API_KEY
• TTS_BASE_URL or TEXTTOSPEECHSKILLS_BASE_URL

The default base URL is https://texttospeechskills.com.

Why MCP Plus Skill?

MCP gives the agent callable tools. The skill gives the agent the workflow: validate markup, estimate credits, choose or create a template, generate, then fetch audio status.

They are different primitives, but users should be able to install them together.