AgamiAI

agami-eval

Runs the golden evaluation harness for a profile. Picks one golden dataset, regenerates SQL for every question in it, executes each statement against the org's own warehouse, scores the result against the confirmed answer key, and reports the verdicts — failures first, with the errored, the unscored and the unconfirmed each kept separate. Scoring is deterministic and happens in agami-core; the skill reports it and never re-judges it. Neither the answer key nor the generated statement is printed: both stay in a local JSON artifact the report points at.

agami-save-golden

Writes golden-dataset items for a profile through two doors. The import door turns a question bank — a CSV, or a table pasted into chat — into items after the parsed rows have been shown and agreed to: a row that already carries a statement is written confirmed, a bare question is written unconfirmed. The save door writes one question, the statement that answered it and the result the person accepted, as a confirmed item. The curation door applies the changes queued on the golden-dataset explorer page, which may weaken a claim and may never grant one. Every write goes through agami-core's writer, is re-read by the runner's own reader before it is kept, and is append-only: a write that would change an item that already exists stops and shows the before and the after. This skill writes only; it never runs or scores a dataset.

agami-reconcile

Reconciles known (label, expected_value) numbers from an existing dashboard against agami's answers. Input can be a SCREENSHOT of a Metabase / Power BI / Tableau / Looker dashboard (Claude's vision extracts the pairs), a CSV, or numbers pasted inline — the user doesn't need to know which; they can just ask. For each pair, the skill generates a matching NL question, runs it through the active profile's semantic model, diffs actual vs expected, and surfaces matches in green and mismatches in red with drill-down receipts. The strongest onboarding demo for a skeptical data engineer — either we agree with their numbers (trust earned via evidence) or we surface a real definitional disagreement (trust earned via transparency).

AI & Automation Listed

agami-connect

End-to-end database connection for agami: sets up credentials on first run (DB-type picker → writes <artifacts_dir>/local/credentials.example for the user to fill in), then introspects the live DB directly into the agami semantic model (subject areas, tables, columns, relationships with join cardinality, deep-table column groups, sensitive-column flags) under <artifacts_dir>/<profile>/. The structural model is built deterministically by the agami-core semantic_model package (catalog mode, or a probe-mode fallback when the catalog is locked down); the skill then layers LLM enrichment (descriptions, entities, metrics) and seeds EXPLAIN-validated NL→SQL examples. Every model write is gated by the semantic-model validator — no breaking model is ever persisted.

DevOps & Infrastructure Listed

agami-deploy

EARLY ACCESS (in testing) — usable today, but newer than the local single-player path; feedback welcome via a GitHub issue. Prepares a ready-to-run, self-hosted agami deploy bundle ON THE USER'S MACHINE so a team can stand up a shareable MCP server their Claude connects to. Conversationally gathers the hard-floor inputs (hostname, admin identity), auto-detects the local model, writes docker-compose.yml + Caddyfile + a filled agami.env (referencing the PUBLISHED image ghcr.io/agamiai/agami-core — no clone, no build), and stages the model artifacts. Generates the signing secret via deploy_preflight; the admin password is typed by the user into the file (never in chat). Then runs `docker compose up` if Docker is local, otherwise prints the exact VM steps + the shareable MCP URL. Username/password auth only on this paved path.

AI & Automation Listed

agami-model

The single dashboard for the active profile's semantic model — browse, curate, AND sign off the trust layer in one surface. Browse every subject area, table, field, metric, entity, and join with live search; edit descriptions/metrics/entities/joins; exclude tables/columns you don't want queried; add new metrics; edit datasource.md. Its **Review tab** is the trust-layer sign-off queue: approve / reject the AI-proposed metrics (Rule 1 — a query using an unsigned metric still answers but carries a warning until it's approved), entities, and inferred joins (Rule 2 — lazy, usable while unreviewed). Every action is queued, submitted back to Claude as one feedback block, applied via the curation engine, and gated by the validator before it touches the YAML. (This skill absorbed the former `/agami-review`.)

agami-query

Answers natural-language questions about the user's database. Loads the agami semantic model (subject areas, tables, columns, relationships with join cardinality, entities, metrics) and few-shot examples from <artifacts_dir>/<profile>/, generates SQL via the examples-first traversal (pick subject area → match examples → resolve entities/metrics → compound table context), executes it locally via the user's chosen tool (psql / mysql / snowsql / sqlite3 native CLI, DuckDB binary, or the Python driver `execute_sql.py` — which runs the scope gates and reports fan-trap/chasm-trap and aggregation findings on the receipt), returns results as a markdown table with optional CSV export, and renders Chart.js HTML charts on request. All execution is local — no data leaves the machine.

AI & Automation Listed

agami-save-correction

Saves a user correction so future queries learn from it. Always appends a (question, corrected_sql) pair to the subject area's example library under <artifacts_dir>/<profile>/prompt_examples/<area>/. Additionally, classifies the correction and — when applicable — applies a surgical edit to the semantic model itself (relationship fix, column metadata, or new metric) via the curation engine. Every model edit is validated before write; the validator is the binding gate, and a failed validation reverts. Shows the user a model diff for approval before any model mutation.

API & Backend Listed

agami-serve

Wires the local agami MCP server (python -m mcp_harness) into the Claude Desktop app in one step, so you can ask your database questions from Claude Desktop — not just inside Claude Code. Auto-detects the right Python interpreter (the one with your DB driver), installs the agami-core package into it so the registration survives plugin updates, and safely merges the entry into claude_desktop_config.json (backup + atomic write, preserving every other key). The local server is the mirror of the hosted Agami connector — same tools, local backend — so this is also how a developer feels the exact experience their business end-users would get.