No spin. AskLedger is a design-partner-stage company building an independent record for enterprise AI, so these answers describe how the system is built to work and where its limits are. If something is a claim rather than proof, we say so.
No. The verifier is open and runs without us. It checks the signature, the hash-chain link and the timestamp entirely on the receipt itself, with no login and no call to any AskLedger server. A regulator, auditor or customer can run the same open tool, and it keeps working even if AskLedger disappears. That independence is the point: it's what makes a receipt evidence rather than a claim.
A receipt proves that a known signer recorded specific fields, and that those fields haven't changed since. It also shows how linked events and checkpoints relate, which policies were evaluated, and which costs and outcomes the platform connected.
It does not automatically prove:
In short: a receipt is strong evidence about the record, not a guarantee about the world. We're deliberate about that line.
No, and we don't try to replace either. Gateways handle routing and reliability; observability tools handle tracing and debugging. AskLedger sits above them, adding identity, policy, cost, outcomes and independent cryptographic proof, and joining it all into one cross-stack record. It's designed to integrate with tools like Portkey, Langfuse, Helicone and OpenTelemetry, importing their traces and costs, rather than competing to out-route them.
By default, no: we work from metadata and hashes, not raw prompts and responses. Capturing actual content is policy-controlled: you decide by role, use case, geography and data classification whether and where content is ever stored. Where content is captured, field-level encryption and selective disclosure keep it sealed, and prompt visibility is exposed only under approved roles and investigations. The design principle is to preserve proof without collecting unnecessary content, and to keep employee risk signals separate from workplace surveillance.
With deterministic, traceable math, not black-box estimates. Every recommendation is based on your real usage, pricing and policy data, and follows a closed loop: baseline (normalized cost before any change), implement (the config, routing or licence change, linked to a signed event), and measure and verify (an agreed post-change period compared against that baseline). We publish each realized saving with its assumptions, exclusions and evidence references. An AI may summarize the findings, but the underlying calculation is always reproducible. Our north-star is verified customer savings, not recommendations generated or "potential" savings estimated.
Yes, for the open pieces. The SDK, the verifier and the local usage-and-cost view are free and open source under Apache-2.0. You can instrument an app, generate and verify receipts, and inspect a basic cost view without buying anything. The enterprise platform is paid: that's where cross-system intelligence, discovery, organization-wide workflows, retention, compliance mappings and support live. Signing and verifying receipts are, and are intended to remain, free and unlimited.
Yes, it's on the roadmap for enterprise, with region-specific storage, customer-managed encryption keys and air-gapped deployment options. Because we're at the design-partner stage, on-prem is delivered as part of enterprise engagements rather than a self-serve download today. The open-source SDK, verifier and local view already run entirely in your own environment.
Two ways, depending on how you like to evaluate things:
No, and we won't claim it does. AskLedger supports auditability and the evidence obligations found in frameworks like the NIST AI RMF, the EU AI Act and ISO/IEC 42001, by producing signed, portable records, framework mappings and evidence packages formatted so an auditor can check them. But these frameworks don't mandate any particular receipt technology, certification is organizational rather than a product feature, and we never promise automatic legal compliance. We help you demonstrate what happened; whether that satisfies a given obligation depends on your use case and your counsel.
We're building this with a small number of design partners who have real AI spend and production applications. Some capabilities described across the site are on the roadmap rather than generally available today. When you talk to us, we'll be clear about what's live, what's in progress, and what a pilot would actually cover.
The fastest way to judge an accountability tool is to verify a receipt and see that it checks out without us. Start there, or book the diagnostic to see value against your own spend.