Vantage RuntimeAI · Adopt

Editor integration

Install runtimeai-ide from PyPI — forecast cost, match library scenarios, or generate a custom check-ride from the SQL, DDL, or policy file in your editor. No Vantage monorepo clone. No PYTHONPATH.

pip install runtimeai-ide →Wire Cursor MCP ↓

Install in two minutes

Works in any repo — dbt, notebooks, agent configs. RuntimeAI talks to the hosted API at vantageai.cc; you only bring a provider key when you want live check-rides. 

pip install runtimeai-ide
python3 -m runtimeai_ide doctor
python3 -m runtimeai_ide suggest --file models/orders.sql
python3 -m runtimeai_ide forecast --scenario de_sql_optimization_v1 --turns 8
python3 -m runtimeai_ide preflight --question "Which model is good enough for our SQL agent?" --file models/orders.sql

Package on PyPI · source on GitHub.

Cursor MCP

Add to ~/.cursor/mcp.json or your project .cursor/mcp.json. Restart Cursor → Settings → MCP → runtimeai should be green. 

{
  "mcpServers": {
    "runtimeai": {
      "command": "python3",
      "args": ["-m", "runtimeai_ide", "mcp"],
      "env": {
        "RUNTIMEAI_BASE_URL": "https://www.vantageai.cc",
        "OPENROUTER_API_KEY": "${env:OPENROUTER_API_KEY}"
      }
    }
  }
}

First Agent prompt (open any SQL file): “Use runtimeai_forecast_cost for de_sql_optimization_v1 with 8 turns, then suggest a scenario for my open SQL file.”

Add your own scenario (same file): “Use runtimeai_generate_scenario on my open SQL file — draft a partition-pruning check-ride from the attached schema.” Save the JSON, then reference custom_… in CI — see Add scenarios guide.

Add a custom scenario from your repo

Library scenarios (e.g. de_sql_optimization_v1) cover common patterns. When your tables, policies, or APIs are specific to your stack, generate a draft from context — no hand-written YAML. 

runtimeai-ide generate \
  --file models/fact_orders.sql \
  --brief "Agent must fix partition pruning using attached schema" \
  --out runtimeai-draft.json

In Cursor Agent chat, call runtimeai_generate_scenario instead. Works without OPENROUTER_API_KEY (hosted generation; server credits apply).

With vs without OPENROUTER_API_KEY

Planning and scenario design are free. Bring a key only when you want to execute and score an agent.

No key required

Explore & plan

  • runtimeai_suggest_scenario — map SQL/brief → library scenario (local)
  • runtimeai_forecast_cost — per-model eval cost estimate (hosted API)
  • runtimeai_generate_scenario — draft custom check-ride (hosted API; server credits)
  • runtimeai_doctor — API check + capability summary

Key + vantage-core

Execute & gate

  • runtimeai_run_checkride — live eval with pass/fail rubric
  • Requires pip install vantage-core and OPENROUTER_API_KEY
  • Same scenarios you gate in CI/CD

MCP tools & CLI

  1. runtimeai_suggest_scenario — pick a library scenario from a file path or brief
  2. runtimeai_forecast_cost — estimate per-run cost before spending token budget
  3. runtimeai_generate_scenario — draft JSON from DDL, SQL, policy, or OpenAPI context
  4. runtimeai_preflight_topic — open question → scenario picks, forecast, and clarity brief (same flow as Preflight)
  5. runtimeai_run_checkride — run vantage-core locally (key required)
  6. runtimeai_doctor — verify API reachability and explain what works without a key
For search engines and LLMs
runtimeai-ide: pip install MCP + CLI for Cursor. Forecast, suggest, generate without OPENROUTER_API_KEY; run_checkride needs vantage-core and provider key.