Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
104 commits
Select commit Hold shift + click to select a range
3916f06
feat: `/savings`: Databricks savings check over altimate-code (1.8.0)
shreyastelkar Nov 17, 2026
f6985dd
feat: `/savings`: Draft the fix, a band on Databricks config, and rev…
shreyastelkar Nov 17, 2026
1c91372
feat: `/savings`: clusters and jobs, a what-if row, and what a change…
shreyastelkar Nov 18, 2026
a4ada13
fix: make the Draft the fix request say where to edit, else what to w…
shreyastelkar Nov 18, 2026
4320e23
fix: /savings review fixes: honest verdicts, the band, the window's l…
shreyastelkar Nov 18, 2026
ae99032
fix: /savings: a dearer serverless choice says so; Databricks-only of…
shreyastelkar Nov 18, 2026
5ae1bf1
feat: /savings across every workspace the connection sees
shreyastelkar Nov 18, 2026
14d2581
fix: an altimate-code without `tool run` gets "update", not "failed: …
shreyastelkar Nov 18, 2026
a002118
fix: count a workspace that only a finding is in
shreyastelkar Nov 18, 2026
fd623ab
docs: README and CHANGELOG: every workspace, old CLIs
shreyastelkar Nov 18, 2026
29ad1a9
fix: /savings runs altimate-code in assessment mode
shreyastelkar Nov 18, 2026
0f172da
fix: what-if and saved-since lines state the gap between the amounts …
shreyastelkar Nov 18, 2026
4b8d543
fix: Draft the fix names CLI commands that work and change only the f…
shreyastelkar Nov 18, 2026
9478abb
fix: a small gap between large amounts stays to the cent; jobs draft …
shreyastelkar Nov 18, 2026
0a94105
fix: the jobs draft starts from the job as it is and resends each cha…
shreyastelkar Nov 18, 2026
ddab79b
Merge main into feat/savings-via-altimate-code
shreyastelkar Nov 18, 2026
dd76bf4
fix: /savings tells an old altimate-code by its whole usage, not the …
shreyastelkar Nov 18, 2026
bbe3310
refactor: keep explain's doc comment on explain; read the metadata once
shreyastelkar Nov 18, 2026
38066b2
fix: /savings asks the CLI's version first and never starts an older …
shreyastelkar Nov 18, 2026
5f52bdb
docs: PRIVACY.md says serverless SQL warehouses get the over-sized check
shreyastelkar Nov 18, 2026
81fb9d7
fix: /savings refuses a source build: `local` does not show assessmen…
shreyastelkar Nov 18, 2026
1d7aafd
feat: /savings gives the go-ahead the savings tool asks for, only fro…
shreyastelkar Nov 18, 2026
a857c45
fix: a thrown /savings run keeps its connection in the terminal fallback
shreyastelkar Nov 18, 2026
a118391
fix: /savings shows a job's host cluster, task runs, hours and DBUs
shreyastelkar Nov 18, 2026
23e2dba
fix: a control character in a workspace name no longer blanks the /sa…
shreyastelkar Nov 18, 2026
306bb25
Merge origin/main into feat/savings-via-altimate-code
shreyastelkar Nov 19, 2026
91821ce
fix: /savings shows a resized warehouse's amounts as observed spend, …
shreyastelkar Nov 19, 2026
73b55b6
fix: /savings counts every workspace the check read, not only those w…
shreyastelkar Nov 19, 2026
783030d
fix: /savings job lines never print NaN, "(undefined)", empty quotes …
shreyastelkar Nov 20, 2026
886f57a
fix: /savings says nothing was checked when no check could run
shreyastelkar Nov 20, 2026
c8c0393
fix: /savings reads the CLI version as the status line does, and says…
shreyastelkar Nov 20, 2026
e84c38a
fix: the /savings offer fires only on a databricks command the shell …
shreyastelkar Nov 20, 2026
30f6b39
fix: /savings keeps a resized-warehouse baseline per workspace
shreyastelkar Nov 20, 2026
02c1ba4
fix: `runTool` sends a call with an environment or `noDaemon` through…
shreyastelkar Nov 20, 2026
4f5cd9b
refactor: tidy /savings and show the observed-spend note once
shreyastelkar Nov 20, 2026
353b8fa
fix: /savings evidence never prints NaN, Run again starts one run, an…
shreyastelkar Nov 20, 2026
9f83ff7
fix: the skill points at /savings, and the band's Run starts one run
shreyastelkar Nov 20, 2026
29e58b7
fix: /savings reads the shell line and the version probe as the shell…
shreyastelkar Nov 20, 2026
a2810dd
fix: /savings reads past every wrapper before the CLI, and never fail…
shreyastelkar Nov 20, 2026
44f5934
fix: /savings drops a what-if row left with no choices
shreyastelkar Nov 20, 2026
eabf501
fix: a heredoc's body in a Bash line ends at its own terminator
shreyastelkar Nov 20, 2026
4294de4
fix: /savings draws an answer missing a field instead of "$NaN" or a …
shreyastelkar Nov 20, 2026
386da1b
fix: /savings no longer reads a heredoc line, and keeps a report whos…
shreyastelkar Nov 20, 2026
21596f3
fix: /savings draws a finding without a title with a plain one
shreyastelkar Nov 20, 2026
2b5928a
fix: /savings reads a Bash line with the shared shell reader
shreyastelkar Nov 20, 2026
5ce8bb8
refactor: share the shell reader between the push check and /savings
shreyastelkar Nov 20, 2026
ca44a26
fix: /savings reads xargs values and shell -c scripts in a Bash line
shreyastelkar Nov 20, 2026
b60c294
refactor: one rule for the script a shell runs from -c
shreyastelkar Nov 20, 2026
4fba945
fix: a shell option cluster ending in o takes the next word as its value
shreyastelkar Nov 20, 2026
29ae296
Merge origin/main into feat/savings-via-altimate-code
shreyastelkar Nov 20, 2026
90ec299
fix: each o in a shell option cluster takes a value; -c without a scr…
shreyastelkar Nov 20, 2026
f5867bc
fix: an unknown shell option before -c no longer hides the script fro…
shreyastelkar Nov 20, 2026
b2c09da
fix: the -c fallback uses the push check's own pattern
shreyastelkar Nov 20, 2026
e749896
fix: shellScript reads the shell's own words, without redirections
shreyastelkar Nov 20, 2026
2fb5dce
Revert "fix: shellScript reads the shell's own words, without redirec…
shreyastelkar Nov 20, 2026
605f61e
Revert "fix: the -c fallback uses the push check's own pattern"
shreyastelkar Nov 20, 2026
24825c0
Revert "fix: an unknown shell option before -c no longer hides the sc…
shreyastelkar Nov 20, 2026
8212734
Revert the shell option-cluster part of "fix: each o in a shell optio…
shreyastelkar Nov 20, 2026
1ccb810
fix: the push check reads -c as main does; a /clear drops a /savings …
shreyastelkar Nov 20, 2026
e15f923
fix: a Run again press beaten by a /clear is dropped before it marks …
shreyastelkar Nov 20, 2026
6b4f911
fix: a band press beaten by a /clear opens no pane and never marks th…
shreyastelkar Nov 20, 2026
a35bb98
fix: /savings takes its generation before it waits, and its doc names…
shreyastelkar Nov 20, 2026
36ad2fd
fix: every step of a /savings run checks it is still current before i…
shreyastelkar Nov 20, 2026
f1d0d16
fix: the running flag is written only while its run is current, check…
shreyastelkar Nov 20, 2026
34e210d
fix: a /clear during the running write or the pane open leaves no pan…
shreyastelkar Nov 20, 2026
2ee9f05
fix: a what-if choice opens only its own warehouse when two workspace…
shreyastelkar Nov 20, 2026
9d18744
fix: /savings names the scoped install package, cleans CLI error text…
shreyastelkar Nov 20, 2026
fb2a259
fix: the spend-since-a-change line names its workspace and leaves out…
shreyastelkar Nov 20, 2026
53d3875
fix: the band does not pin a name flagged in several workspaces on on…
shreyastelkar Nov 20, 2026
5339672
fix: a size read on one side only is no resize, and the band quotes n…
shreyastelkar Nov 20, 2026
9332ff8
refactor: a kept baseline holds each warehouse setting as read, merge…
shreyastelkar Nov 20, 2026
ab0faec
fix: a baseline fills an unknown setting whatever else changed, and a…
shreyastelkar Nov 20, 2026
a474758
fix: a warehouse cost sent as a numeric string is read as a cost
shreyastelkar Nov 20, 2026
7390ff9
fix: fill an unknown setting only from a read with no change; keep /c…
shreyastelkar Nov 20, 2026
8e3cd0b
fix: keep a baseline as one run's read; guard writes by conversation,…
shreyastelkar Nov 20, 2026
e2d27ae
fix: narrow bundle detection, one window length, unnamed band matches…
shreyastelkar Nov 20, 2026
54c7142
fix: read bundle compute by its YAML block, not a pattern
shreyastelkar Nov 20, 2026
dca6c07
fix: decide Databricks config by the file first; read a string lookba…
shreyastelkar Nov 20, 2026
de40bb6
fix: accept an unquoted Terraform label; pin the bundle block's child…
shreyastelkar Nov 20, 2026
56647c7
fix: a stale press cannot free the Run guard; read a CLI error as tex…
shreyastelkar Nov 20, 2026
8cb28e1
fix: read a CLI error object as its reason; skip a Terraform block co…
shreyastelkar Nov 20, 2026
af5c137
fix: an empty CLI error shows the CLI's own text; guard the Hide pres…
shreyastelkar Nov 20, 2026
2e3fe8c
fix: the band's Run press clears the band only for its own conversati…
shreyastelkar Nov 20, 2026
2f1e9c1
test: a Hide press just before /clear leaves the new conversation's o…
shreyastelkar Nov 20, 2026
a41a312
fix: type the baseline inside its cost guard; read Terraform by its c…
shreyastelkar Nov 20, 2026
bee1f23
fix: keep a block comment's line breaks when reading Terraform; make …
shreyastelkar Nov 20, 2026
c19422b
fix: the band reads a Terraform file's code, not its comments, for co…
shreyastelkar Nov 20, 2026
65b868b
fix: read a dashed heredoc delimiter; make the quoted-string Terrafor…
shreyastelkar Nov 20, 2026
bed357a
fix: read Terraform by line, not by parsing; make the Hide-before-cle…
shreyastelkar Nov 20, 2026
138a666
fix: a Terraform file starting with a byte-order mark is read; README…
shreyastelkar Nov 20, 2026
6be4309
fix: Hide has one conversation guard, the press-level one its test re…
shreyastelkar Nov 20, 2026
c12f9e6
fix: Hide keeps its apply-time conversation check, tested by a /clear…
shreyastelkar Nov 20, 2026
4f2c652
fix: /savings typed while a band press starts its run says the run is…
shreyastelkar Nov 20, 2026
bf27d26
test: /savings typed while a band press runs its check to the end say…
shreyastelkar Nov 20, 2026
1c357da
fix: a running flag /savings read before a /clear does not open the p…
shreyastelkar Nov 20, 2026
0892cf4
fix: /savings and its Run presses start a check from one claim, not f…
shreyastelkar Nov 20, 2026
c861bc4
fix: a new session lets go of an old run's claim; a bare /savings dur…
shreyastelkar Nov 20, 2026
08571f9
test: a check from before a /clear that ends after it does not free t…
shreyastelkar Nov 20, 2026
de3e39b
fix: a new session keeps a run's claim: the module that holds it star…
shreyastelkar Nov 20, 2026
f791000
fix: /savings that finds a check running re-checks after its pane ope…
shreyastelkar Nov 20, 2026
350d43b
fix: a run's claim is let go even when its last write fails; "just fi…
shreyastelkar Nov 20, 2026
f2f9802
fix: a Run press whose running-flag write fails lets its claim go, as…
shreyastelkar Nov 20, 2026
6894cb3
refactor: one helper lets go of a run's claim, for its end and for a …
shreyastelkar Nov 20, 2026
31b536a
test: the dbt artifact-race test waits for each call's read to begin …
shreyastelkar Nov 21, 2026
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion .claude-plugin/marketplace.json
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,7 @@
},
"metadata": {
"description": "Altimate's Claude Code plugin — data tools through the altimate-code MCP server, or the model-free CLI where MCP is disallowed, plus delegation of long dbt jobs to the CLI agent.",
"version": "1.8.3"
"version": "1.9.0"
},
"plugins": [
{
Expand Down
9 changes: 8 additions & 1 deletion PRIVACY.md
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,7 @@ This policy supplements the Altimate AI [Privacy Policy](https://www.altimate.ai
- When a task **is** delegated, your task text is passed to the altimate-code CLI, which sends it to an LLM provider to plan and execute the work — the Altimate LLM Gateway by default, or a provider you configure (BYOK).
- Warehouse queries, dbt commands, and file edits run **locally on your machine, under your credentials**. Your warehouse credentials are never sent to Altimate.
- In a project linked to an Altimate workspace, the plugin copies the workspace's team skills into the project's `.claude/skills/` and refreshes them from Altimate in the background at session start and at most hourly. Their text loads into the Claude Code conversation like any project skill. `ALTIMATE_CLAUDE_SKILL_SYNC=0` turns this off and removes the copies at the next session start.
- The plugin counts its own use anonymously, through the altimate-code CLI's usage telemetry: one event at each session start (plugin and altimate-code versions, OS, a random install id) and one each time a panel or command opens or a gate holds or allows a command (its name and outcome). No session id, project, model, column or file names, SQL or session text, and nothing that links them to your account or to the CLI's other telemetry. `ALTIMATE_TELEMETRY_DISABLED=true` turns it off. See [Plugin usage events](#plugin-usage-events).
- The plugin counts its own use anonymously, through the altimate-code CLI's usage telemetry: one event at each session start (plugin and altimate-code versions, OS, a random install id) and one each time a panel or command opens or a gate holds or allows a command (its name and outcome); `/savings` sends none. No session id, project, model, column or file names, SQL or session text, and nothing that links them to your account or to the CLI's other telemetry. `ALTIMATE_TELEMETRY_DISABLED=true` turns it off. See [Plugin usage events](#plugin-usage-events).
- Installing the plugin is an explicit act — `/plugin marketplace add` does not install it, and `/plugin install` never auto-runs. Nothing is delegated until the plugin is installed and you send a task that matches the skill.

## What the plugin does and does not do
Expand Down Expand Up @@ -122,6 +122,7 @@ To count installs, active use and which features are used, the plugin asks the l
- No event contains project, model, column or file names, SQL, query results or session text. The CLI accepts each field only from a fixed list or pattern (a version, a UUID, one of the names above) and refuses anything else.
- **Turning it off:** `ALTIMATE_TELEMETRY_DISABLED=true` (or `1`) in the environment Claude Code starts in: the plugin then starts no CLI for these events, or for the lineage report's, and makes no install id. The CLI also drops them under its own opt-out (`ALTIMATE_TELEMETRY_DISABLED`, or `telemetry.disabled` in its config), as for every other event.
- The events are sent in the background and never delay a hook. An altimate-code that predates them refuses them, and nothing is sent or shown.
- **`/savings` sends no usage event.** Typing it, pressing its buttons and opening its pane are not counted, and nothing about the check (its queries, findings or amounts) goes to Altimate. See [The Databricks savings check](#the-databricks-savings-check).

## Data flows when a task is delegated

Expand All @@ -134,6 +135,10 @@ To count installs, active use and which features are used, the plugin asks the l

**Bring Your Own Key (BYOK):** if you configure your own provider (Anthropic, OpenAI, Bedrock, Azure OpenAI, Ollama, and others), your prompts go directly from your machine to that provider. Altimate is not in that delegation path; workspace service requests still go to Altimate.

## The Databricks savings check

`/savings` runs altimate-code's `finops_databricks_savings` tool on your machine, with your own Databricks connection. It reads usage from your workspace's system tables and decides the savings locally. The plugin runs it with telemetry turned off, in altimate-code's assessment mode (`ALTIMATE_ASSESSMENT_MODE=true`) and outside its tool daemon. In assessment mode the CLI sends no usage event, not even the one-time telemetry opt-out event, and does not fetch the model catalog or download ripgrep, so your warehouse is the only host the run contacts; an altimate-code release without assessment mode ignores the setting and would still send that opt-out event and make those fetches, so the plugin first asks `altimate --version` (which sends nothing) and does not start a release older than 0.12.12, or a source build (which prints `local` whatever its age), for the check. The plugin sends no usage event of its own for `/savings` either (see [Plugin usage events](#plugin-usage-events)), so nothing about the check, its queries or its results goes to Altimate. The findings are shown in a pane, not added to the conversation. Not assessed: Databricks Apps spend; a serverless SQL warehouse gets only the over-sized check. The run starts the connection's SQL warehouse if it is stopped, and needs the `@databricks/sql` driver installed. Costs cover the six full billed days before the window's last day, scaled to seven, which read about 7% below the seven billed days on one measured workspace. To show a flagged warehouse's observed spend before and after a later change, the plugin keeps its size, type and weekly cost in Claude Code's local plugin store on your machine. **Draft the fix** puts one finding's details in your prompt box; they reach your model provider only if you send that prompt. When Claude runs the same tool itself, its answer is a tool result like any other and reaches your model provider.

## Usage telemetry

The altimate-code CLI emits anonymous product usage telemetry: event names, tool categories, token counts, timing, and error types. Telemetry never includes your code, queries, credentials, file paths, or prompt content, and error messages pass through a redaction layer (API keys, emails, and internal hostnames are masked). Events are keyed to an anonymous machine ID — a random UUID generated on first run and stored at `~/.altimate/machine-id`, not tied to your hardware or identity. For logged-in users, the email identifier is SHA-256 hashed before sending. The full event list is published in the [Telemetry reference](https://docs.altimate.sh/reference/telemetry/).
Expand Down Expand Up @@ -173,6 +178,8 @@ The plugin and the altimate-code CLI are professional developer tools, not direc

We will update this document in this repository and revise the effective date when our practices change. Material changes to Altimate-wide practices are governed by the master [Privacy Policy](https://www.altimate.ai/privacy).

**November 19, 2026:** added [The Databricks savings check](#the-databricks-savings-check) (plugin 1.9.0): `/savings` runs altimate-code's savings tool with telemetry off, sends no plugin usage event, and keeps the findings out of the conversation.

**November 19, 2026:** added [Plugin usage events](#plugin-usage-events) (plugin 1.8.2): the anonymous session start and feature-used events, and the plugin's own check of the telemetry opt-out.

**November 17, 2026:** added [dbt helpers](#dbt-helpers) (plugin 1.8.0): the dbt commands the plugin runs in a dbt project, and what it keeps of their results.
Expand Down
2 changes: 1 addition & 1 deletion plugins/altimate-code/.claude-plugin/plugin.json
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
{
"name": "altimate-code",
"displayName": "Altimate Code",
"version": "1.8.3",
"version": "1.9.0",
"description": "Gives Claude Code altimate-code's model-free data tools (column-level lineage, impact analysis, PII detection, data-diff, FinOps, query equivalence, dialect translation, warehouse queries) through its MCP server, or through `altimate tool run` when MCP servers are not allowed, with delegation to the altimate-code agent for long dbt jobs, access to a linked Altimate workspace's knowledge, and the workspace's team rules in each session.",
"author": {
"name": "AltimateAI",
Expand Down
14 changes: 13 additions & 1 deletion plugins/altimate-code/CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,19 @@

All notable changes to this plugin will be documented in this file.

## [Unreleased]
## [1.9.0] — 2026-11-19

### Added
- `/savings [connection]`: a Databricks savings check beside the chat, over altimate-code's `finops_databricks_savings` tool (four checks: Migrate Classic/Pro to Serverless, Over-Sized Warehouse, Idle Cluster Detection, Job Running on All-Purpose Compute). It runs with your own Databricks connection, in altimate-code's assessment mode (telemetry off and no other contact with Altimate or GitHub: the warehouse is the only host reached; needs an altimate-code release with `ALTIMATE_ASSESSMENT_MODE`) and outside the tool daemon, and shows each finding with its evidence and a monthly range at list price, each resource counted once. Needs an altimate-code release with that tool; with an older one the pane says to update.
- `/savings` gives the savings tool the go-ahead it asks for (`confirm: true`) only when you ran it yourself; a plugin, the SDK or a schedule gets no run. The check stops after 9 minutes, cancelling any query still on the warehouse, and a check that could not run is named with its reason.
- Draft the fix: a finding's change goes into the prompt box for Claude to draft against the repository; nothing is sent until you press Enter.
- What if: each flagged warehouse's cost as serverless and at every smaller size, and whether its cores still fit its queries.
- Spend since a change: a flagged warehouse's size, type and cost are kept on this machine from one run, and a later run that finds the size or type changed shows its observed spend a month before and after. The two amounts come from different weeks, so the pane labels them as observed spend and does not call the difference saved.
- `/savings` sends no plugin usage event: neither the command nor its pane is counted, and its queries and results never go to Altimate.
- A line above the prompt when the session edits Databricks bundle or Terraform files or runs a `databricks` compute command: the flagged resource's saving, or a once-a-session offer of `/savings`.
- One connection checks every workspace the account's system tables show; the pane counts every workspace the CLI read (`workspaces_checked`), one with only clusters or jobs and no finding included, and names each finding's workspace. With a CLI that does not report the count, it counts the workspaces of the warehouses and findings.
- `/savings` asks `altimate --version` first and does not start an altimate-code older than 0.12.12, or a source build, for the check: before assessment mode, its `tool run` would send the one-time telemetry opt-out event.
- `runTool` takes an environment and `noDaemon` for one CLI run, and hands back a CLI's raw output when it is not an answer (an altimate-code too old for the tool gets "update").

### Fixed
- After `/clear`, the new conversation loads the team comments again (the band, and the threads Claude gets when it reads a commented model) instead of waiting for `/comments` or the next poll, and project health keeps its baseline, the changes not yet checked and the issues found (`/health` no longer says the session has no project; a check the `/clear` cut short runs again). The status line keeps the integrations and altimate-code parts; health draws its own again.
Expand Down
14 changes: 13 additions & 1 deletion plugins/altimate-code/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -156,6 +156,18 @@ Compile and validate runs a query on your warehouse after every model edit (`dbt

When Claude mentions `altimate-dbt`, it is the dbt command line that comes with altimate-code (`npm i -g @altimateai/altimate-code`, the same install as the plugin's engine). It finds your dbt project folder, Python environment and warehouse adapter by itself, and altimate-code's agent uses it instead of the raw dbt tool for better error messages. Its commands include `init`, `doctor`, `compile`, `build`, `run`, `test`, `deps`, `columns <model>`, `graph` (lineage) and `execute "<sql>"` (a query through the dbt adapter). You do not need to run it yourself: Claude and the altimate-code agent call it when a task needs it, and the plugin's own dbt helpers above do not depend on it. See [altimate-dbt in the altimate-code docs](https://docs.altimate.sh/data-engineering/tools/dbt-tools/#altimate-dbt-cli).

## Databricks savings check

`/savings [connection]` checks a Databricks workspace for savings with four of Altimate's Databricks checks: Classic or Pro SQL warehouses that would cost less as serverless, warehouses larger than their queries need, idle all-purpose clusters, and scheduled or continuous jobs running on all-purpose clusters. A pane beside the chat shows each finding with what it rests on and a monthly range at list price, each resource counted once at its largest saving.

- **Draft the fix.** Each finding has a button that puts the change in your prompt box (serverless, a smaller `cluster_size`, auto-termination, or a job cluster) for Claude to draft against your bundle or Terraform, or as CLI JSON. Nothing is sent until you press Enter, and nothing is applied.
- **What if.** For each flagged warehouse, press serverless or any smaller size to see what it would cost a month and whether its cores still cover what its queries need.
- **Spend since a change.** The first run that flags a warehouse remembers its size, type and cost on this machine (`$.store`), all from that run. A later run that finds the size or type changed shows its observed spend a month before and after the change, and names only the settings read both times. A setting the CLI did not list then is taken from the first later run that reads every setting the baseline knew, finds none changed and gives a cost: that run's read, date and spend replace the first run's, so the comparison starts from it. The two amounts come from different weeks, so a quieter or busier week moves them too: they are what was billed, not a measured saving.
- **At the right moment.** Editing a bundle's `databricks.yml`, a YAML file under `resources/` whose top-level `resources:` block lists jobs, pipelines, clusters or SQL warehouses, or a Terraform file with a line that starts with a `databricks` provider, resource or data source, or running a `databricks` warehouses, clusters, bundle or jobs command, shows a line above the prompt: the flagged resource's saving when the file names it, else, when the file mentions warehouses or clusters (a command always counts), a once-a-session offer of `/savings`. Nothing runs by itself.
- **What runs.** altimate-code's `finops_databricks_savings` tool, with your own altimate-code Databricks connection (the first one that runs through a SQL warehouse, or the one named). It reads seven days of Databricks system tables (`system.billing`, `system.compute`, `system.query`, `system.lakeflow`) on that SQL warehouse, which Databricks starts if it is stopped and bills while it runs, so it runs only when you type `/savings` (or press its Run button), never on its own; the check stops after 9 minutes, cancelling any query still running. System tables cover every workspace of the account they can see, so one connection checks all of them, and the pane names each finding's workspace. A check that could not run (its tables could not be read, or the time ran out) is named as not checked, with the reason; the others still report.
- **What it needs.** altimate-code 0.12.12 or later, which has that tool and assessment mode, and a Databricks connection whose `http_path` is a SQL warehouse (`/sql/1.0/warehouses/<id>`) with an access token that can read those system tables.
- **What leaves your machine.** The plugin first asks `altimate --version`, which starts no telemetry, and does not start an older altimate-code, or a source build (which prints `local` whatever its age), for the check. It then runs the tool with `ALTIMATE_TELEMETRY_DISABLED=true`, `ALTIMATE_ASSESSMENT_MODE=true` and `--no-daemon`, so neither altimate-code nor altimate-core sends usage telemetry for it. `/savings` sends no plugin usage event either: typing it and opening its pane are not counted, and the check's queries and results never go to Altimate. The findings stay in the pane until you send a Draft the fix request. Claude can also run the same tool itself; its answer then reaches your model provider like any other tool result.

## Settings

The plugin has seven settings, shown when you enable it (`userConfig` in `.claude-plugin/plugin.json`). All are optional: Claude Code may report them as "not set", and each then takes the default below. `dbt_mods` is used by the dbt helpers, `maxScanBytes` by the scan check (BigQuery, Snowflake, Databricks), `healthBlocksPush` and `healthConfigPath` by the project health check, the two comments settings by team comments, and `show_edit_impact` by the per-edit impact note.
Expand Down Expand Up @@ -185,7 +197,7 @@ MCP tools and `tool run` run locally under your credentials; tool results Claude

The dbt helpers run the project's dbt on this machine. What they add to Claude's context (dbt's errors, column names, the manifest's facts about a model) reaches its model provider like any tool result. The rows of the Query results pane are kept on this machine only, in Claude Code's plugin store, until you clear them (bookmarked queries keep their rows through **Clear history**, and turning `dbt_mods` off does not delete them; see PRIVACY.md).

The plugin counts its own use anonymously through the altimate-code CLI's usage telemetry (`altimate-code tool report-event`; no network connection of the plugin's own): `plugin_session_start` at each session start, with the plugin and altimate-code versions, the OS and a random install id kept in the plugin's Claude Code store; and `plugin_feature_used` when `/lineage`, `/dbt-show`, `/profile-ctes`, `/scan-cost`, `/comments`, `/sql-graph`, `/health`, `/runs` or `/to-model` opens, or the cost check or the health push gate holds or allows a command, with the feature's name and the outcome. No session id, project, model, column or file names, SQL or session text, and none of the identifiers altimate-code's other telemetry carries (account, machine id). Set `ALTIMATE_TELEMETRY_DISABLED=true` to turn it off (with altimate-code's other telemetry). An altimate-code that predates these events drops them silently. Every event and field is listed in [PRIVACY.md](../../PRIVACY.md#plugin-usage-events).
The plugin counts its own use anonymously through the altimate-code CLI's usage telemetry (`altimate-code tool report-event`; no network connection of the plugin's own): `plugin_session_start` at each session start, with the plugin and altimate-code versions, the OS and a random install id kept in the plugin's Claude Code store; and `plugin_feature_used` when `/lineage`, `/dbt-show`, `/profile-ctes`, `/scan-cost`, `/comments`, `/sql-graph`, `/health`, `/runs` or `/to-model` opens, or the cost check or the health push gate holds or allows a command, with the feature's name and the outcome. `/savings` is not counted: it sends no usage event. No session id, project, model, column or file names, SQL or session text, and none of the identifiers altimate-code's other telemetry carries (account, machine id). Set `ALTIMATE_TELEMETRY_DISABLED=true` to turn it off (with altimate-code's other telemetry). An altimate-code that predates these events drops them silently. Every event and field is listed in [PRIVACY.md](../../PRIVACY.md#plugin-usage-events).

Delegating a task with `altimate run` passes your prompt text to the altimate-code CLI, which sends it to your configured LLM provider — the Altimate LLM Gateway by default, or your own provider (BYOK), in which case Altimate never sees it. Warehouse credentials and query results stay on your machine.

Expand Down
Loading
Loading