Documentation
TelltaleMCP tools
Updated
Telltale has no chart dashboard. You analyse your app by asking an agent, which calls the tools on Telltale's MCP server. Every tool is scoped to one app, takes that app's numeric app_id, and returns bounded JSON.
Connect your agent
The simplest route is through Scalpels. Install the Scalpels MCP server in your agent and create a token with the Call installed apps ability; see Use Scalpels from your agent. Your agent can then list the MCP-enabled products your team runs, including Telltale, and relay calls to it. Each call carries your identity and role: team Owners, Admins and Members can all use Telltale's tools. How that relay works is described under Console.
Telltale's MCP server is served at /mcp on your deployment, with its destructive tools on a separate /mcp/destructive endpoint. Telltale does not have a screen for minting its own MCP tokens.
Then ask in plain language, for example:
Which errors are new since version 2.3.0 of our mobile app, and how many installs hit them?
What share of active installs is on the latest release?
Finding your app
| Tool | What it returns |
|---|---|
apps |
Up to 100 apps with their id, name, latest retained event, versions live in the last 30 days and cumulative client drops. Start here to get the app_id. |
ingest_health |
Client drops, known rejections, rate-limit hits and recently seen client versions over days (1 to 90, default 30). |
Usage and releases
| Tool | What it returns |
|---|---|
active_users |
DAU, WAU and MAU as of a UTC date (as_of, YYYY-MM-DD, default today), with version and platform breakdowns. |
release_adoption |
Daily active-install counts and shares by app version over days (1 to 90). |
event_counts |
Event or screen counts (metric: event or screen) over days, optionally broken down by version, platform, os or locale. |
retention |
First-seen-week cohorts and weekly returning-install retention over weeks (0 to 12, default 8). |
These read durable daily aggregates, so they keep working after raw events are pruned.
Behaviour
| Tool | What it returns |
|---|---|
screen_flow |
The most common screen-to-screen transitions and session entry and exit screens over days, up to limit (1 to 50). |
funnel |
Ordered per-install conversion and drop-off for 1 to 10 event names in steps, within window_minutes (default 1440, up to 10080) of the first step, over days. |
install_timeline |
Up to 200 retained events, in order, for exactly one install_id or one session_id. |
These read retained raw events, so they can only look back as far as the retention window (90 days by default; see Configuration).
sessions returns session count, average, median and 95th-percentile length, and average screens per session over days. It reads session records, which the raw-event prune does not delete.
Errors
| Tool | What it returns |
|---|---|
errors |
Error groups seen in the last days (default 30), up to limit (1 to 50), with occurrences, installs affected, first and last seen times and versions. Pass new_since_version to flag groups that first appeared after that version. |
error_detail |
One group by its 64-character fingerprint (from errors), with its totals, a bounded sample error with stack data and the distribution of context it occurred in. |
Errors are grouped by fingerprint. These are PHP errors reported through Laravel, not native crashes.
Administration tools
Three tools change data. Each is two-phase: the first call returns a preview of what would happen, and nothing changes until it is confirmed with a single-use confirmation.
| Tool | What it does |
|---|---|
revoke_install |
Revokes one install's token so that device can no longer send or re-register. Its history is kept. |
delete_install |
Deletes one install with its raw events and sessions. The preview reports how many would go. Daily aggregates are preserved. |
rotate_ingest_key |
Invalidates an app's ingest value and returns the replacement exactly once, like the dashboard's Rotate ingest value. |
See Privacy and data for when to revoke and when to delete.