{
  "markdown": "# Sprout MCP\n\n**Drop your Opus bill 80%.** Model-tiered content pipeline for MCP — cheap models seed work, expensive models verify it.\n\nSprout routes tasks to the right model tier automatically. Haiku drafts, Sonnet fact-checks, Opus verifies. Every chunk tracks provenance, confidence, and cost.\n\n## Install\n\n```bash\nuvx sprout-mcp\n```\n\nOr add to Claude Code's MCP config (`~/.claude/settings.json`):\n\n```json\n{\n  \"mcpServers\": {\n    \"sprout\": {\n      \"command\": \"uvx\",\n      \"args\": [\"sprout-mcp\"]\n    }\n  }\n}\n```\n\n## How It Works\n\n```\n  Haiku (seed)  →  Sonnet (watered)  →  Opus (sprouted)\n    Draft            Fact-check           Verify\n    $0.005/M         $0.015/M             $0.075/M\n```\n\n1. **Seed** — Haiku drafts content cheaply (summarization, extraction, first passes)\n2. **Water** — Sonnet reviews and fact-checks the seeds\n3. **Sprout** — Opus deep-verifies only what passed Sonnet's review\n\nInstead of running everything through Opus at $75/M output tokens, most work stays at Haiku's $5/M. Only the final verification — typically 10-20% of total work — touches Opus.\n\n## Tools (13)\n\n| Tool | Description |\n|------|-------------|\n| `submit_chunk` | Store content with provenance (model, task type, sources) |\n| `get_review_queue` | List chunks needing review, filtered by confidence/project |\n| `mark_reviewed` | Promote (seed→watered→sprouted) or reject chunks |\n| `recommend_model` | Get model recommendation for a task type |\n| `get_stats` | Dashboard of chunk counts, confidence levels, token usage |\n| `export_chunks` | Export verified chunks as JSON |\n| `opus_test` | Generate structured review summary for batch verification |\n| `schedule_task` | Schedule tasks to run at a specific time or delay |\n| `list_scheduled` | View pending scheduled tasks |\n| `cancel_scheduled` | Cancel a pending scheduled task |\n| `configure_routing` | Add/update routing rules at runtime |\n| `get_cost_report` | Estimated spend per model with real pricing |\n| `retry_on_error` | Track failed attempts with backoff guidance |\n\n## Configuration\n\n### Environment Variables\n\n| Variable | Default | Description |\n|----------|---------|-------------|\n| `SPROUT_DB_PATH` | `~/.sprout/sprout.db` | SQLite database location |\n| `SPROUT_CONFIG` | *(none)* | Path to JSON config file for custom routes and pricing |\n| `SPROUT_MAX_RETRIES` | `3` | Max retry attempts before giving up |\n| `SPROUT_RETRY_BACKOFF` | `2.0` | Exponential backoff base (seconds) |\n\n### Custom Config File\n\nCreate a JSON file and point `SPROUT_CONFIG` at it:\n\n```json\n{\n  \"routes\": {\n    \"code_review\": { \"tier\": \"sonnet\", \"reason\": \"Code analysis needs reasoning\" },\n    \"translation\": { \"tier\": \"haiku\", \"reason\": \"Straightforward language task\" }\n  },\n  \"pricing\": {\n    \"custom-model\": 10.00\n  }\n}\n```\n\n### Default Routing Table\n\n| Task Type | Tier | Why |\n|-----------|------|-----|\n| `biography_synthesis` | haiku | Factual summarization |\n| `council_description` | haiku | Historical summarization |\n| `document_synopsis` | haiku | Content summarization |\n| `json_validation` | haiku | Structural verification |\n| `summarization` | haiku | General summarization |\n| `data_extraction` | haiku | Structured extraction |\n| `fact_check_first_pass` | sonnet | Cross-reference claims |\n| `code_review` | sonnet | Code analysis |\n| `fact_check_final` | opus | Deep factual verification |\n| `theological_analysis` | opus | Domain expertise required |\n| `complex_analysis` | opus | Deep reasoning required |\n\nUnknown task types default to haiku — start cheap, escalate if needed.\n\n## Example Workflow\n\n```\nYou: Use recommend_model for \"biography_synthesis\"\nSprout: biography_synthesis → haiku-4.5 (Factual summarization)\n\nYou: Use submit_chunk to store the Haiku output\nSprout: Stored chunk abc12345 [seed] for person-001.biography\n\nYou: Use get_review_queue to see what needs fact-checking\nSprout: 1 chunk pending review\n\nYou: Use mark_reviewed to promote after Sonnet fact-checks it\nSprout: Chunk abc12345 → watered (verified by sonnet-4.6)\n\nYou: Use get_cost_report\nSprout: haiku-4.5: ~1,300 tokens (1 chunk) — $0.0065\n        Total: $0.0065\n```\n\n## Development\n\n```bash\ngit clone https://github.com/mepsopti/sprout-mcp.git\ncd sprout-mcp\nuv sync --extra dev\nuv run pytest\n```\n\n## Support\n\nIf Sprout saves you money on your AI bill, consider buying me a coffee:\n\n[![Buy Me A Coffee](https://img.shields.io/badge/Buy%20Me%20A%20Coffee-support-yellow?style=flat&logo=buy-me-a-coffee)](https://buymeacoffee.com/mepsopti)\n\n## License\n\nMIT\n\n`mcp-name: io.github.mepsopti/sprout-mcp`\n",
  "bytes": 4547,
  "sha": "2ca668830e07b975ca6148b21d66e994746699bd37288d902dc9d7cbb6999f0c",
  "repo_slug": "mepsopti/sprout-mcp",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/mcp_io_github_mepsopti_sprout_mcp_5cfee8ee/readme"
}