Sherlock Calls
for ClickHouse
Your ClickHouse warehouse holds every call event, session record, and operational metric your stack produces. But querying it means SQL, access credentials, and context switching. Sherlock changes that: ask any question about your ClickHouse data in plain English from Slack and get a sourced answer in under 5 seconds. Invite-only beta for teams running call and event analytics on ClickHouse.
TL;DR — What beta users get access to
- 1
Sherlock Calls connects to ClickHouse via host, port, database, username, and password — read-only access to your event and analytics tables in under 2 minutes, no SQL required.
- 2
Ask 'How many calls failed in the last 24 hours by error type?' or 'What is the p95 latency for ElevenLabs TTS events this week?' — Sherlock translates to SQL, queries ClickHouse, and returns a sourced answer in Slack.
- 3
ClickHouse is the fastest open-source columnar database for real-time analytics — purpose-built for the event volumes that voice AI and telephony stacks generate at scale.
<5s
Time to query ClickHouse event data from Slack
0
SQL required — ask in plain English
Any
ClickHouse table or view Sherlock can query
About ClickHouse
What is ClickHouse?
ClickHouse is an open-source columnar database management system designed for real-time analytical processing of large-scale event data. It is optimized for high-throughput INSERT workloads and sub-second analytical queries across billions of rows, making it the standard choice for teams storing call events, session logs, webhook payloads, and operational metrics at scale. ClickHouse supports SQL, provides a native HTTP interface, and is available as a self-hosted deployment or as ClickHouse Cloud. It is widely used in voice AI and telephony stacks for storing CDRs, TTS latency events, call failure logs, and aggregated usage metrics.
About Sherlock Calls
What is Sherlock Calls?
Sherlock Calls is an AI call detective that lives in your Slack workspace. It connects to voice providers, CRMs, and ad platforms, then answers any question about your calls in plain English — in under 5 seconds, with no code, no new dashboards, and no SDK required. Think of it as a senior analyst always on call, inside the tools your team already uses.
Beta Setup
Connect ClickHouse to Sherlock in 2 minutes
No code, no webhooks, no new dashboards. Beta users get direct onboarding support.
- 1
Provide your ClickHouse connection details
Enter your ClickHouse host, port, database name, username, and password in Sherlock's integration settings. Sherlock validates the connection with a test query before saving. Read-only credentials are recommended — Sherlock never writes to your database.
- 2
Describe your schema (optional but recommended)
Optionally describe your key tables and columns in plain English so Sherlock can generate more accurate queries. For example: 'call_events table has call_id, status, duration_ms, error_code, provider, created_at'. The more context you provide, the better the query accuracy.
- 3
Start asking questions in Slack
Ask 'How many calls failed with error 20003 in the last 7 days?' or 'Show me the p99 TTS latency by hour for the past 24 hours' — Sherlock translates to ClickHouse SQL, executes the query, and returns structured results in your Slack thread.
Beta Use Cases
What beta teams are asking Sherlock about ClickHouse
- SC“How many calls failed in the last 24 hours broken down by error code?”
- SC“What is the p95 and p99 TTS latency for ElevenLabs events this week?”
- SC“Show me call volume by hour for the past 7 days”
- SC“Which call campaigns have the lowest completion rate this month?”
- SC“How many unique callers did we have yesterday?”
- SC“What is the average call duration by provider for the past 30 days?”
Beta Features
What beta users can do with ClickHouse
Natural Language to ClickHouse SQL
Ask any question about your call and event data in plain English. Sherlock translates to accurate ClickHouse SQL, executes it, and returns the result with the query shown for transparency.
Real-Time Event Investigation
Query billions of call events, TTS latency records, and webhook payloads in real time. ClickHouse's columnar architecture means sub-second answers even on large datasets.
Cross-Provider Correlation
Combine ClickHouse query results with data from Twilio, ElevenLabs, and other connected providers in a single investigation. Correlate your raw event warehouse with live provider APIs.
Schema-Aware Query Generation
Sherlock learns your table structure and generates queries that respect your column names, data types, and partitioning strategy — not generic SQL that fails on your actual schema.
Read-Only and Secure
Sherlock connects with read-only credentials and never executes INSERT, UPDATE, or DELETE statements. Your data warehouse is always protected.
AI Tools
5 specialized tools, zero configuration
Sherlock picks the right tool automatically. You just ask in plain English — the AI knows which APIs to call and how to correlate the results.
Beta users get direct access to all tools below — with personal onboarding from the Sherlock team.
FAQ
Common questions about Sherlock + ClickHouse
How does Sherlock Calls connect to ClickHouse?
- Sherlock connects via the ClickHouse HTTP interface using your host, port, database, username, and password. Read-only credentials are strongly recommended. The connection is validated with a test query before being saved.
Does Sherlock work with ClickHouse Cloud as well as self-hosted?
- Yes — Sherlock connects to any ClickHouse instance accessible via HTTP or HTTPS, including ClickHouse Cloud, self-hosted deployments, and managed ClickHouse on AWS, GCP, or Azure.
Can Sherlock query any table in my ClickHouse database?
- Sherlock can query any table or view that the connected user has SELECT access to. We recommend creating a dedicated read-only user with access scoped to the specific tables relevant to your call and event analytics.
Does Sherlock store my ClickHouse data?
- Sherlock does not store your ClickHouse table data. Query results are used to generate the Slack response and are not persisted. Your ClickHouse credentials are encrypted at rest. Full details at usesherlock.ai/privacy.
What kinds of questions can I ask about my ClickHouse data?
- Any analytical question your data supports: failure counts by error code, latency percentiles, call volume trends, cost aggregations, provider comparisons, and anomaly detection. If it can be answered with a SQL SELECT, Sherlock can answer it in plain English.
Apply for early access to Sherlock + ClickHouse
We're accepting a select group of beta users to shape the ClickHouse integration. Tell us about your stack and we'll reach out personally if you're a fit.
Even more powerful combined
Also works with