LeanBuddy

LeanBuddy is the AI assistant built into the LeanSignal app. It answers questions about your tenant’s telemetry — grounded in what your agents actually collect and what your demands actually store — and links you straight to the dashboard, logs, metrics, or traces view that shows the answer.

Open it with the LeanBuddy button on the right edge of the app. The panel docks on the right and stays put while you navigate, so you can follow its links without losing the conversation. Close it and the transcript survives — reopening picks up where you left off.

What you can ask

LeanBuddy queries the same stores the app itself reads, through read-only tools — it can look, but it cannot change anything. Good starting points (the panel suggests these when a conversation is empty):

• Triage — “Give me an overview of errors and spikes in the last 24h.” LeanBuddy sweeps stored logs and metrics for anomalies and summarizes what stands out.
• Agent health — “Is my agent healthy?” Connection status, version, and whether the demand set is current.
• What’s collectable — “Which workloads are available on the agent?” Compares what the agent sees at the edge with what is demanded centrally.
• Demand hygiene — “Analyze my demands and suggest improvements.” Finds filters that store nothing, overlap, or cost more than they are used.
• Incidents — describe a symptom (“checkout latency doubled around 14:00”) and LeanBuddy narrows it down with targeted queries.

Answers come with link buttons that open the relevant dashboard, logs, metrics, or traces page. When LeanBuddy drafts a query for you, an Open prefilled button seeds the matching explore page with it — you run, refine, and save it yourself; LeanBuddy never demands telemetry on its own.

Because central storage holds only the demanded subset, absence of data is itself an answer: LeanBuddy will tell you when something is not demanded (rather than not happening) and point at the demand gap.

Conversations

Every conversation is saved. The history button in the panel header lists them; open one to continue it, or delete it permanently. New chat starts a fresh conversation — do this when you change topic, since a shorter conversation is also a cheaper one (see budgets below).

The tenant notebook

The notebook button in the panel header opens a shared, tenant-wide scratchpad of facts LeanBuddy carries into every conversation — service names, naming conventions, known quirks (“staging is expected to be noisy on Sundays”). Ask LeanBuddy to remember something and it writes a note there; editors and admins can edit or prune the notebook directly. It is capped at 6 KiB — keep it to durable facts, not conversation history.

Credits and the daily allowance

LeanBuddy is metered in credits (1 credit = 1 US cent of model spend), with a hard daily allowance per user and per tenant that resets at midnight UTC. The meter under the panel header shows today’s usage; each answer shows what it cost and how many telemetry queries it ran. When the allowance runs out, LeanBuddy stops until the reset — there is no overage billing, by design.

The full cost model — credits, cost-weighted tokens, default allowances — is published at LeanBuddy budgets.

Availability

LeanBuddy is provided by the platform — there is no API key to bring and nothing to configure in your tenant. If the panel says LeanBuddy is not available, it has not been enabled for your tenant yet; open a case under Settings → Support and we will switch it on.

For IDEs and agents: the MCP server

Everything LeanBuddy can see is also exposed over an MCP server, so your own AI tooling (Claude Code, IDE assistants) can query the tenant’s telemetry directly. Personal access tokens are managed on the Preferences page — see AI access (MCP) for creating a token and connecting each client. MCP access does not draw on LeanBuddy’s credit allowance — your agent brings its own model, so the spend happens in your own tooling.