XorqXorq
./waitlist

Verifiable AI for Enterprise Data Teams

A shared catalog and multi-engine execution layer for governed AI on enterprise data. Run any model, trace every result, and reuse verified work on your existing warehouse.

AI generates numbers.Xorq verifies them.

talk to us
> what was Q3 revenue for the West region?
without xorq · a guess ≈ $24M? which script? · no lineage
nobody signed it
with xorq · a proof $0.0M SELECT sum(amount) … region='West'
@7be9896e · J. Alvarez · 4 hops

Xorq is The Agentic Execution System for AI:

01Every step runs as a real query on your warehouse.
02Every result is human-approved.
03Every answer is reused by the agents that follow.
Verified Reporting Reports and decks where every number traces to a real query, regenerated on demand. explore Agent Verification A sub-agent your assistants call for lineage-backed answers, so any output is checkable at the source. explore Decision Agents Agents that build and run models for pricing, risk, and recommendations inside your environment, every step verified. explore
Snowflake Databricks DuckDB Postgres Apache DataFusion 20+ more
01How it works

Xorq gets agentic work to production.

customer agents
Coding agents
Chat & copilots
AI data analysts
intent ↓↑ result
XorqXorq xorq execution system
Executes every step is a real query
Catalogs work lands as reusable artifacts
Traces lineage back to blessed sources
Caches the same work never runs twice
Gates a human signs off before it ships
real queries ↓↑ rows
Warehouses
Object stores
Filesystems
customer data
Any model, any lab, any harness
Human interpretable
Your VPC, your keys. Nothing phones home
xorq desktop client
02Proof

A smaller model. Better accuracy. Half the tokens.

dabstep·public financial industry benchmark

An AI agent answers 450 analytics questions from a real payments database.

Haikuon its own
50%
Sonneton its own
75%
Haiku + Xorqverified
84%
same model · same prompt · the only change is xorq
+34 pts accuracy · 50% → 84%
60% fewer tokens 566k → 228k per task
14 → 6 turns less back-and-forth per answer
03What compounds

Every question your agents answer becomes an asset.

Xorq Ledger

A conversation ends and its work goes with it. The Ledger keeps what an agent did as a versioned, traceable artifact your teams and systems can find, verify, and rerun.

xorq ledger · finance
catalog · finance
revenue_by_region
revenue_west_q3
churn_cohorts_2026
dispute_rate_daily
pricing_elasticity
ltv_by_segment
revenue_west_q3 @7be9896e · J. Alvarez · 2 days ago
verified ✓
result$23.8M
runs142
cacheHIT
expression sales.filter(region == 'West', quarter == 'Q3').agg(sum(amount)) → SELECT sum(amount) FROM sales WHERE region='West' AND quarter='Q3'
lineage · 4 hops
raw.payments→ stg.sales→ dim.region→ revenue_west_q3
history
approved by M. Chen
@7be9896e · 2d ago
committed by J. Alvarez
@7be9896e · 2d ago
rerun by finance-agent
@31c0a4f2 · 9d ago
created by reporting-agent
@a91d7e03 · 6w ago
Recorded Every run lands in the Ledger as a named, searchable artifact, not a chat transcript.
Versioned Each change gets a commit. Roll back, diff, or approve like code.
Traceable Lineage from any number back to the source tables it came from.
Reusable The next agent, or the next analyst, starts from verified work instead of a blank prompt.
04Runs where you do

Start on a laptop. Scale to your VPC.

One platform, three ways to run it. Bring the Xorq harness or plug Xorq into the agents you already use. Same Ledger, same proofs, wherever it lands.

land · one person
Desktop One person, their keys and warehouse. Local first, with the Xorq harness built in.
✓Xorq harness included ✓Runs locally · DuckDB or your warehouse ✓Personal Ledger
macOS free to start
grow · shared
Teams Shared workspaces, dashboards, collaboration. Run the Desktop client, or go headless and call Xorq from the agents you already use.
✓Desktop or headless ✓Xorq harness or Claude plugin ✓Shared Ledger · approvals · lineage
hosted by xorq claude plugin
expand · platform
Enterprise Headless in your VPC. Your keys, your network, your governance. Nothing phones home.
✓Headless · any harness ✓SSO · audit · policy gates ✓Ledger inside your perimeter
your vpc talk to us
one
platform
Xorq Ledgerbusiness definitions, verified and runnable
Xorq Cachework runs once, reused not recomputed
Xorq Anywhereone expression, every engine
05The difference

Trust by execution - not another observation layer.

xorq the execution system that runs the work itself, as real queries, so the answer is the proof.
not a chat window that hands you an answer you cannot check.
not a log or a trace that describes what an agent did after the fact.
not a second model grading the first one, or a static dashboard scoring its answers.
// ready when you are

Give your agents work the business can stand behind.

Talk to us or explore the open-source library →

Things people ask first.

What is Xorq Desktop?+

A native desktop app. Point it at your data, pick a workflow, and an agent works against the Xorq Ledger, a shared store of governed, executable expressions.

Do I need to migrate my data?+

No. Connects to Snowflake, Databricks, DuckDB, S3, Postgres, whatever you already run.

Do I need to set up an MCP server?+

No. Desktop packages the Ledger, verification gates, caching, and warehouse connections; you configure it with your existing credentials.

How does this get through security review?+

It runs on the analyst’s machine with your existing warehouse credentials and LLM subscription. Bulk data stays out of a new vendor cloud, reducing the scope of security review.

What is the Xorq Ledger?+

An executable catalog of what your agents have run: expression metadata, lineage, and cached results, addressed by hash. Metric definitions live here as runnable code; agents compose new entries on verified work.

How is this different from a notebook + an LLM?+

Notebooks accumulate output. Xorq accumulates reusable work. Last week’s expression remains addressable, so the next agent can build on it instead of starting over.

Two people, same question. Conflicts?+

Same expression, same hash, automatic dedup. Identity is content, not file path.

Linux / Windows?+

Desktop is Mac-first. The library and Textual TUI run everywhere Python does.