A calmer operating surface for AI sessions, traffic, and cost drift.
TknScope records the evidence you actually need during review: which workflow ran, where traffic went, how sessions grouped, and when the bill started moving.
Sessions indexed
12,480
Provider events / day
84k
Spend reviewed
$18.4k
Workspace digest
March review baseline
codex-cli / nanoclaw-admin
Grouped session29 turns rolled into one review surface
117M tokens
nanoclaw / claude-code
Live ingestSessionEnd hooks landing in production
5 sessions
laptop proxy / provider events
Streaming trafficOpenAI, Gemini, Anthropic, OpenRouter
Traffic view
Traffic
24h
provider call ledger
Sessions
Grouped
conversation-first review
Auth
org-bound workspace access
Built for teams that need to explain what happened, not just chart it.
The product surface stays narrow on purpose: sessions, traffic, alerts, imports, keys, and shared review workflows.
Grouped session review
Collapse fragmented Codex task rows back into the conversation they came from so the review surface matches the work that actually happened.
Conversation-first, not event-first.
Provider traffic ledger
Track proxy traffic, process origin, credential fingerprints, and provider metadata in one timeline instead of reconstructing it from shell history.
Built for operational attribution.
Cost drift visibility
Watch usage totals, provider mix, and workflow changes move together before a billing alert shows up after the fact.
The budget view should be forensic.
Workspace controls
Keep analysts, owners, and operators in the same org with role-based access instead of spreading review across local state and ad hoc screenshots.
Review paths stay shared.
Historical ingest
Backfill existing session archives and imports when you need continuity instead of forcing adoption to start from a blank dashboard.
Useful during migration.
API-key ingestion
Push data directly from agents, hooks, and batch workflows without rebuilding your stack around a single SDK.
Works with real systems.
Capture real work, not synthetic metrics.
Start by installing the Claude Code hook or post directly to the ingest API. TknScope then turns that raw session flow into dashboards, filters, and review queues.
Recent activity
Claude Code hook installed
claude-code
Session ingest accepted
api/ingest
Cost delta crossed threshold
alert rule
Analyst acknowledged session
workspace activity
Expected result
Set the path once, then let the product accumulate the history you need.
Create the workspace
Start with the org boundary first so sign-in, API keys, and review roles all point at the same ledger.
Route real traffic
Install the hook, enable the proxy path, or post session archives so production-like data shows up immediately.
Review grouped evidence
Use session, traffic, and alert views together so expensive or abnormal runs can be explained with actual artifacts.
Operator checklist
- 1
Bind auth and org
Google sign-in, password login, and ingest keys should all terminate inside the same workspace.
- 2
Verify the routing
Make sure sessions are grouped correctly and provider events land before you trust totals.
- 3
Establish a baseline
Once the ledger is stable, alerts and spend conversations stop depending on memory.
Start with the hook. Expand with the API and proxy path.
TknScope is most useful when it becomes the path of record. Begin with the existing hook, then wire direct ingest and provider traffic so the ledger stays complete.
Claude Code
Best for automatic session capture
Install once, let Stop events send session summaries automatically, then review grouped conversations in the workspace.
REST + proxy
Best for custom agents and provider visibility
Post your own message history through the ingest API and route provider traffic through the proxy when you need attribution beyond transcripts.
settings.json
POST /api/ingest
Make TknScope the path of record before the next usage question arrives.
Start with a workspace, install the hook, route the traffic you care about, and let the product accumulate the operating history you wish you already had.