[hkaltai]

Do not merge — just experiments. This branch exists only to evaluate how Claude Code (and other agents) should call altimate-code data tools directly, over MCP or from the shell.

What this experiment contains

• Brings in altimate-code mcp serve from feat(mcp): altimate-code mcp serve exposes data tools to MCP clients without an LLM #1543 and the workspace tools from feat(mcp): expose workspace context to coding clients #1576.
• One adapter, two front doors: mcp serve and a new model-free tool CLI (tool list | schema | run | context | daemon) execute tools through the same adapter, so both share compact schemas, sql_execute with tables and query_from, result spill files, answer_commit, and read-only enforcement.
• mcp serve default set: the five feat(mcp): altimate-code mcp serve exposes data tools to MCP clients without an LLM #1543 tools plus impact_analysis, dbt_lineage, altimate_core_column_lineage, schema_detect_pii, data_diff, altimate_core_equivalence, sql_translate; --all-tools exposes the full 70-tool allow-list.
• tool context prints the analyst method and connection/table listing that mcp serve sends at startup, for shell agents.
• Long-running per-project process (auto-start, idle exit, version- and connections-scoped socket) keeps connections warm and releases DuckDB/SQLite file locks between calls.
• --connections / ALTIMATE_TOOL_CONNECTIONS, --args - from stdin, --workspace-tools.
• DROP DATABASE, DROP SCHEMA and TRUNCATE stay blocked even with --allow-write.
• link --claude-md (opt-in) writes CLI routing guidance to CLAUDE.md for clients that cannot use MCP.
• The CI change from feat(mcp): altimate-code mcp serve exposes data tools to MCP clients without an LLM #1543 is intentionally not included.

Validation

• bun run typecheck clean; test/altimate/tool-cli + test/cli/help 50/50; ALTIMATE_MCP_E2E=1 test/altimate/mcp-server 101/101; other feat(mcp): altimate-code mcp serve exposes data tools to MCP clients without an LLM #1543/feat(mcp): expose workspace context to coding clients #1576 test files 761/761.
• End to end: tool context, tool run via the daemon and in-process, and mcp serve over stdio (initialize, tools/list, tools/call) against DuckDB.
• Companion plugin experiment: [DO NOT MERGE] Experiment: MCP-first plugin with a CLI fallback altimate-claude-plugin#11. Benchmark results are shared internally.

🤖 Generated with Claude Code

---

Summary by cubic

Experimental branch, not for merge: evaluates how external coding agents call altimate-code data tools directly, via a new model-free tool CLI or mcp serve over stdio.

• tool list | schema | run | context | daemon and mcp serve execute the same adapter, so both share compact schemas, sql_execute options, result spill files, answer_commit, and read-only enforcement.
• A per-project daemon keeps connections and the schema cache warm between calls and releases DuckDB/SQLite file locks.
• mcp serve defaults to the original five tools plus impact, lineage, PII, data-diff, equivalence and translation tools; --all-tools opens the full 70-tool allow-list.
• MCP requests are forced read-only; DROP DATABASE, DROP SCHEMA and TRUNCATE stay blocked even with --allow-write.
• link --claude-md writes CLI routing guidance into CLAUDE.md for clients that cannot use MCP.

Behavior changes

• schema_inspect lists schemas and up to 500 tables with catalog row estimates when table is omitted; MongoDB catalog discovery is bounded and closes cursors early.
• answer_commit accepts asked and readings arguments, and its rejection checks moved into a dedicated module with per-turn limits.
• The analyst now requests high reasoning effort on adaptive-thinking models unless the user or agent config sets one.

Validation: typecheck clean; test/altimate/tool-cli + test/cli/help 50/50; ALTIMATE_MCP_E2E=1 test/altimate/mcp-server 101/101; other affected test files 761/761.

^{Written for commit 18e72ea. Summary will update on new commits.}

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

This PR doesn't fully meet our contributing guidelines and PR template.

What needs to be fixed:

• PR description is missing required template sections. Please use the PR template.
• "Cost side effects" section is missing or empty. Describe any model calls, queries or cloud resources whose cost grows with user data, or write "None".

Please edit this PR description to address the above within 2 hours, or it will be automatically closed.

If you believe this was flagged incorrectly, please let a maintainer know.

[github-actions]

Hey! Your PR title [DO NOT MERGE] Experiment: model-free tool CLI and MCP server on one adapter doesn't follow conventional commit format.

Please update it to start with one of:

• feat: or feat(scope): new feature
• fix: or fix(scope): bug fix
• docs: or docs(scope): documentation changes
• chore: or chore(scope): maintenance tasks
• refactor: or refactor(scope): code refactoring
• test: or test(scope): adding or updating tests

Where scope is the package name (e.g., app, desktop, opencode).

See CONTRIBUTING.md for details.

[github-actions]

👋 This PR was automatically closed by our quality checks.

Common reasons:

• New GitHub account with limited contribution history
• PR description doesn't meet our guidelines
• Contribution appears to be AI-generated without meaningful review

If you believe this was a mistake, please open an issue explaining your intended contribution and a maintainer will help you.

[github-actions]

👋 This PR was automatically closed by our quality checks.

Common reasons:

• New GitHub account with limited contribution history
• PR description doesn't meet our guidelines
• Contribution appears to be AI-generated without meaningful review

If you believe this was a mistake, please open an issue explaining your intended contribution and a maintainer will help you.