01 / Collect
One place for the spend.
Import provider CSVs, record events with the JavaScript or Python SDK, or let agents book spend through MCP.
See the integrations →Bookkeeping for AI agents
Your agents spend.
Keep the books.
Bring every token, tool call, and agent purchase into one clear ledger. Know who spent it, why, and whether the books still add up.
No signup to explore · Your demo data stays in your browser
veritally
/ ledgerSample books
A clear trail for every dollar
8 entries. 3 agents.
Token usage and payments, in the same books.
research-agent / R&DAI Services
outreach-agent / GrowthAI Growth
ops-agent / OpsAgent Purchases
Every entry connected.Change one amount. The
chain tells the story.
Bring your existing stack
OpenAIAnthropicGeminiLiteLLMMCP
From agent run to journal entry
Less chasing charges.
More knowing where they went.
A practical accounting layer for the people building agents and the people responsible for their spend.
02 / Attribute
Give each dollar an owner.
Connect spend to an agent, a human principal, and a policy. Work through missing attribution and check budgets before spending.
Work the exceptions queue →03 / Reconcile
Books you can check.
Verify the hash chain, post corrections as adjustments, close the month, and export journal CSVs for QuickBooks or Xero.
Read the verification rule →Start with the spend you already have.
The Accountability Check finds attribution gaps in a usage export. Your file stays on your device.
A closer look
Watch the books come together.
Two short screen recordings of the working product, using sample data. No account required.
01 / The product tour
From unknown spend to clear books.
Explore the dashboard, attribute a missing owner, and export the journal.
Read the transcript
The dashboard opens with sample spend, attribution, budget, and anomaly views. An unowned project appears in the exceptions queue. Assign an agent, principal, and policy to resolve it. Verify the ledger, then download the journal CSV for review.
02 / Follow the proof
One changed number. A visible break.
Verify a ledger, edit a dollar amount, and see exactly where the chain breaks.
Read the transcript
Verify the sample ledger to confirm its eight entries match their stored hashes. Simulate changing an amount without updating its hash. Run verification again: the first altered entry is identified and the following rows are highlighted. Restore the amount and verify that the chain is valid again.
Try the proof yourself
A ledger with nothing to hide.
These eight sample entries are hashed in your browser. Verify them, then change an amount and check again.
| Seq | Date | Agent | GL account | Amount | Hash |
|---|
Each hash commits to the previous hash and the journal entry. Verification detects changes against the stored chain.
A hash chain detects tampering. Independently published chain heads help detect a rewritten chain. How signing and anchoring work →
Corrections belong in new adjusting entries. The demo edits above deliberately bypass that rule to show how verification detects a change.
Public proof pages & badges →Room for your fleet to grow
Start exploring. Grow with confidence.
Try the demo freely. API access and paid plans include onboarding with your team.
Developer
$0 / month
For exploring your first integration.
- REST API, MCP & SDKs
- One tenant and its ledger
- Attribution & budget checks
- Hash-chain verification
Pro
$49 / month
For bringing agent spend into the books.
- Everything in Developer
- Token and payment events
- Monthly close & adjustments
- QuickBooks / Xero journal CSVs
- Weekly spend digest
Team
$199 / month
For a fleet with people behind it.
- Everything in Pro
- Multiple agents & projects
- Priority support
- Personal onboarding call
- Direct roadmap input
Building with agents in production? Apply to the design-partner program. First five teams: free for a year.
Good questions
A little clarity
before you start.
What can I try without an account?
The dashboard runs the ledger engine on sample data in your browser. You can import a CSV, work through attribution, verify entries, and export a journal. The Accountability Check also runs locally.
Does Veritally read my prompts?
The API records structured spend fields such as model, tokens, cost, and attribution. Prompt and completion content are not part of the spend event schema. Local demo files stay in your browser.
Does verification prove every charge is correct?
Verification checks that entries match their stored hash chain. It does not validate the original invoice or prove business purpose. Signed, independently published chain heads add protection against an operator rewriting the chain.
Can I connect this to my accounting system?
Export journal CSVs formatted for QuickBooks or Xero, or use a general double-entry journal export. Review the entries with your accountant before importing them.
Is this tax or compliance advice?
No. Categories and readiness scores help organize a review. Your accountant or legal adviser should assess tax treatment and your particular obligations.
Build the next chapter with us
Your agents.
Your people.
Your books.
We’re onboarding a small group of teams running agents in production. Get a direct line to the builder and help shape the product.
Built by an Enrolled Agent.
Designed for the people
accountable for the spend.