Opportunities › Reliable SQL generation for data analytics agents › Proposed app
Metric Ledger
A context layer that makes AI agents write correct SQL from defined metrics
- Score
- 43
- confidence 47
- Evidence
- 8
- items on the page
- Rivals
- 3
- named in sources
- First version
- 45 d
- estimate
- Type
- Desktop app
- PWA: runs in the browser, installs on Mac and Windows
- Market since
- 2025
- New market · AI estimate
ScreensConcept
Desktop app: Data and business analysts define metrics in YAML, review join planning, and audit query trails in sustained technical work requiring keyboard input and schema review. Sample screens on a computer, in the proposed brand. Their numbers are sample data.
Agent request
Compiled and checkedNet revenue
About this appAI proposal
Metric Ledger sits between your AI agent and your warehouse. You define measures, dimensions, filters and joins once in YAML, with the cardinality of each join. The agent asks for 'net revenue by channel' and the planner compiles the SQL, instead of the agent guessing joins from raw tables.
It also reads your Markdown wikis and Notion pages, so rules such as 'exclude test orders' or 'attribution is last non-direct click' are attached to the metrics they affect. The planner flags fanout and chasm joins before a query runs and tells the agent why.
It is for analysts and teams running Claude Code, Codex or an internal data agent against BigQuery, Snowflake or Postgres, who want answers they can trust without maintaining a heavy BI semantic layer.
- Tagline
- Define the number once. Every agent gets it right.
- Voice
- PreciseCalmPlain
- Why this name
- A ledger is a trusted record where each figure is accounted for. It fits metrics that agents can look up and trace.
In the first version
Members say which features they would use. It tells PWAS what people want; it changes no score.
YAML metric definitions
Write measures, dimensions, filters and joins in plain YAML with cardinality. Versioned in git, readable by analysts.
No member yet
Join planner
Detects fanout and chasm joins and rewrites to pre-aggregate, or refuses the query with a clear reason.
No member yet
Business knowledge linking
Connect Markdown and Notion docs. Rules are attached to the metrics they affect and shown to the agent.
No member yet
Agent-ready endpoint
Agents call named metrics through a tool interface and get compiled SQL, results and the rules applied.
No member yet
Brand identityAI proposal
Check the name against trademark registers before using it.
Palette
#1F5D50Ledger Green · primary#F6F3ECPaper · surface#1B2421Ink · text#E0A526Amber · accent#B5482FBrick · warning
Typography
Aa
Metric Ledger
ABCDEFGHIJKLMNOPQRSTUVWXYZ 0123456789
Define the number once. Every agent gets it right.
PricingAI proposal
The material gives no competitor prices. These tiers are a starting point to test, based on the per-warehouse and usage-based ideas in the opportunity.
Starter
Free
- 1 warehouse
- 25 metrics
- 1,000 agent queries a month
Team
$149 a month
- 2 warehouses
- Unlimited metrics
- 50,000 agent queries a month
- Notion and Markdown sync
- Audit trail
Business
$499 a month
- Unlimited warehouses
- 500,000 agent queries a month
- Drift alerts
- SSO
- Priority support
What it could be worthEstimate
What a software business making this much sells for today, from real deals. Pick a monthly revenue.
If it makes
Worth about
$194,000
Sale range $132,000 to $263,000
$120,000 a year in revenue
From 87 software businesses listed and sold on Empire Flippers in the last year, revenue vetted by the broker: they sell for about 19× their monthly revenue, most between 13× and 26×.
Measured on profit, the middle deal is 32× monthly net profit: about 2.7 years of profit.
An estimate, not an offer or financial advice. A real price depends on profit, growth, churn, how much of the work is the owner's, and the buyer.
How it would growAI proposal
How the product would reach new customers through its own use, and how the first ones would be found. A plan to test, not a result.
- Built-in loop
- Partners and integrations. List it as a tool server for Claude Code, Codex and internal agents, and publish an open YAML format and importers for dbt and LookML so analysts find it where they already work.
- Do they become customers?
- Agent users who find it are the analysts and data leads who pay. Test whether individual analysts convert from the free tier to a team plan.
- First customers
- Data leads at 20 to 200 person companies on BigQuery or Snowflake who already run an internal data agent or use Claude Code for analytics.
- Where to reach them
- Hacker News and data communities: Share worked examples of fanout errors and fixes, and invite readers to try their own warehouse.
- dbt community: Offer a dbt importer and post guides on moving existing models to agent-ready metrics.
- Direct outreach: Contact teams that describe building data agents and offer to define their top ten metrics.
- First milestone
- Five paying teams that each route a real agent workflow through ten or more metrics for a month.
Why people already pay for this
- VerifiedHacker News post received 93 points and 37 comments
- AI inferenceAccuracy is the #1 issue when building production-grade data agents.
- AI inferenceAgents generate syntactically valid SQL that is semantically incorrect due to missing context about data quality, business rules, and schema changes.
- Public claimTeams have built production-grade data agents for dozens of companies, indicating active market demand and willingness to invest in custom solutions.
- AI inference
Tech stack
Every proposal builds on the same PWA base, installable from the browser on any phone or computer. Dashed boxes are what this app adds.
- AppNext.js and TypeScript
- PWAManifest and service worker
- UITailwind CSS
- Data and authSupabase Postgres
- PaymentsStripe
- EmailResend
- HostingVercel
- This app · Query plannerSQL parser and compiler libraryNeeded to plan joins, detect fanout and emit BigQuery, Snowflake and Postgres dialects.
- This app · Agent interfaceMCP-style tool serverLets Claude Code, Codex and internal agents call metrics as tools.
RisksAI inference
- TechnicalCorrect join planning across three SQL dialects is hard; a planner bug could return a wrong number with false confidence.
- CommercialTeams may keep using agents with wikis, or existing semantic layers, if writing YAML definitions feels like too much upfront work.
- DataCustomers must grant warehouse access, so security review may slow sales and demands strict read-only handling.
- AIAgents may bypass the layer and write raw SQL, or link wiki text to the wrong metric, so accuracy gains need measuring.
- PlatformAgent tool standards and warehouse vendors may change, or add their own semantic features, narrowing the gap.
Everything marked AI proposal or AI inference was written by a model from the evidence on the opportunity page. The proposed name must be checked against trademark registers before it is used.