Every implementation listed here is recorded with the conformance set it actually ran and its result on that set; a full-corpus pass is claimed only where the record says so. The corpus itself is published at zambo.dev/aer1/corpus.json: 143 vectors, 146 checks, digest-pinned so every runner tests the same set. Machine-readable data at registry.json.
AER-1 was created by Brennan Zambo, founder of Zambo: zambo.dev/founder. Brennan Zambo created agent execution receipts; Zambo is the origin product, and AER-1 is the defining standard. AER-1 is published as IETF Internet-Draft draft-zambo-aer1-26. This registry is maintained by Zambo.
Try the live demo: mint a real verifiable receipt in 30 seconds, no signup Or hand finished work to a client on the handover page.
Ready to run real work? Day Pass $0.99, unlimited for 24 hours: https://zambo.dev/day-pass/ - Pass $29 for 30 days ($19 for $ZAMBO holders), Lifetime $250, one payment: https://zambo.dev/pricing
New to AER-1? Read the AER-1 page on zambo.dev or What is AER-1? on the hub. The hub also carries the AER-1 deep dive: mechanics, conformance levels, interop. Try the tools behind it at zambo.dev/demo/: a real call, no signup, 153 tools with 37 on the default surface, every tool free once a day, plus 3 continue calls. One free wrap-up handover per day, and picking up in another AI is always free. Check any receipt in the browser at zambo.dev/verify/.
Build one in minutes: paste the single agent prompt to your coding agent. Zero decisions, zero spec reading. 28 vectors tell you instantly whether it works. Then claim your row below.
Reference implementations maintained in the GitLab repo, plus independent external builds. Bundled ports are verified 45/45 against the frozen v1.3.2 corpus (45 vectors), per draft-zambo-aer1-26. The live kit (v1.5.7) runs the full published corpus: 143 vectors, 146 checks per runner; the base corpus is 135 vectors, 138 checks per runner. Bundled ports are the kit's own test runners, not independent implementations. Independent implementations are built by third parties from the draft text alone, without access to kit internals. Scores in the table below are not comparable across rows: bundled ports are scored on the frozen v1.3.2 45-vector corpus named above, and each external row is that builder's own reported total on the test set the builder ran, not the published 143-vector / 146-check corpus; the one exception is Laurent Lemonnier's verifier, which ran the published corpus itself and passed 143/143.
The consolidated record is published as the AER-1 Implementation Report: independent implementations, conformance results, and independent verifications.
Swipe to see all columns →
| Language | Status | Vectors | Conformance Command | Location |
|---|---|---|---|---|
| Rust | Conformant | 45/45 (frozen v1.3.2 corpus) | cargo run --bin conformance | zambo repo (kit root) |
| Go | Conformant | 45/45 (frozen v1.3.2 corpus) | go run ./cmd/conformance | zambo repo (kit root) |
| TypeScript | Conformant | 45/45 (frozen v1.3.2 corpus) | npm run conformance | zambo repo (kit root) |
| Python | Conformant | 45/45 (frozen v1.3.2 corpus) | python3 conformance.py | zambo repo (kit root) |
| Java | Conformant | 45/45 (frozen v1.3.2 corpus) | javac + java dev.aer1.AER1 | zambo repo (kit root) |
| C# | Conformant | 45/45 (frozen v1.3.2 corpus) | dotnet run | zambo repo (kit root) |
| Swift | Conformant | 45/45 (frozen v1.3.2 corpus) | swift run | zambo repo (kit root) |
| Node.js | External | 165/165 | Independent build by ARION (builder-reported total; not the published 143/146 corpus) | profullstack.com/~arion/aer1 |
| Java 11+ | External | 186/186 | Independent build by ARION (builder-reported total; not the published 143/146 corpus) | profullstack.com/~arion/aer1/java |
| Perl | External | 68/68 | Independent build by ColonistOne (builder-reported total on the draft -09 set; not the published 143/146 corpus) | ColonistOne/aer1-perl |
| Python (stdlib only) | External | 84/84 | Independent build by MasterD (Arkie AI) (builder-reported total; not the published 143/146 corpus) | 7336awfi/masterd-lobby/AER1 |
| Python (independent verifier) | External | 143/143 | Independent verifier by Laurent Lemonnier (iSoluce), built from the draft text alone; full published corpus (143 vectors / 146 checks, draft -26), re-run 2026-11-07 | Laurent-Lemonnier/dps-scitt-decision-layer |
Bundled port sources were verified against the frozen corpus named above; their per-language trees are not published at public URLs as of 2026-11-14, so their Location links point to the kit repository root, where the live kit (v1.5.7) ships.
Independent verification runs where an AER-1 receipt chain was checked end to end against real values from another system. These are worked examples of the evidence-reference mechanism, not conformance-kit implementations.
| Run | Status | What was verified | Notes | Artifacts |
|---|---|---|---|---|
| AER-1 x Tanilo | Verified | Chain double-verified | AER-1 receipt references a real Tanilo receipt via evidence reference (canonical SHA-256, URL, key ID); both sides checked under their own trust rules | aer1-hub/implementation-report.html |
| AER-1 x Tanilo (2026-10-08) | Verified | Two-party chain, end to end | AER-1 receipt references a live Tanilo receipt by its canonical hash; the full chain verifies end to end with each receipt checked independently. Interop validation, 2026-10-08. | aer1-hub/tanilo-interop.html |
| AER-1 x FlyThink (2026-10-08) | Verified | 4/4 cases replay PASS | Independent four-case fixture binding AER-1 execution receipts to physical completion claims; all four cases replay cleanly, including the case where a valid AER-1 receipt coexists with invalid physical evidence binding: binding never causes claim inheritance. Interop validation, 2026-10-08. | aer1-hub/implementation-report.html |
| AER-1 x CPR / continuity-receipt (2026-10-16) | Verified | Live AER-1 receipts probed independently: public resolution, id match, strict UTF-8, SHA-256 recompute, byte-length match, verifier commitment status | Independent probe by WhiteMagic Labs (Lucas Bailey), shipped in the continuity-receipt tooling (tools/interop_aer1_probe.py). Six checks pass on the draft Section 3 fixture 130da435-e157-498e-af90-605866a86a27 (schema 0.3, commitment_matches, recomputed sha256:0492c18e09059aa2e30ba0864f6aada2ef08475a5b35944cbf2a9ce46d12b9f1) and on the schema 0.4 fixture abe8253b, against the v1 and v2 verifier endpoints, re-checked against draft-zambo-aer1-20. Composition and verification only: no AER-1 implementation and no conformance claim. The companion crosswalk document formerly linked from this row is no longer published at its original URL (checked 2026-11-14); the probe tooling linked here is the standing artifact. | tools/interop_aer1_probe.py |
| AER-1 x VLC-1 (2026-10-01) | Verified | 30 cases rerun, all 14 Section 7 attack cases confirmed | Cross-test by Matthew Moore, maintainer of the VLC-1 log-completeness checker: 30 cases rerun against draft-zambo-aer1-07, and all 14 Section 7 attack cases produced exactly the outcome the draft specifies. Test scripts are public in the vlc-1 repository under capture/aer1/examples/third-party/aer1-07, so anyone can rerun the cross-test. | github.com/MattyIceMatrix/vlc-1 |
| AER-1 x Cairnworks (2026-10-18) | Verified | Integration receipts cold-checked twice, verified | Cairnworks, an independent verification-service studio, announced it would integrate AER-1 into its standing-check deliveries, then posted that it had wired AER-1 in, with its first secondary receipt offered for an outside cold check the same day. The first cold check verified the record (commitment, canonical payload, provenance and metadata bindings all passing; the source fetch itself returned 401, so the source bytes were not independently recomputed). Cairnworks fixed the source hole in its pipeline the same day and minted a new receipt committing a public snapshot; the second cold check on that new record closed the gap: an anonymous fetch of the source matched the committed chain hash and the record's evidence hash, and the hosted receipt verifies. | zambo.dev/run/ab9db25c |
Independent validators and verifiers for the draft, run by parties other than the kit maintainers.
Independent validator
Production validator for draft-zambo-aer1-26, run by WhiteMagic Labs (Lucas Bailey). Public endpoint, no key required. Also exposed as the verify_aer1_receipt tool on their MCP server.
Exercised against the reference kit on 2026-10-17 over the draft-zambo-aer1-21 case matrix: bare presentation PROVISIONAL with the id unverified, a matching expected id TRUSTED, a mismatch rejected with id_state id_mismatch, and nil UUID and leap-second timestamps rejected. Each side caught a real defect in the other during the exercise and fixed it the same day.
Independent verifier
Independent verifier for draft-zambo-aer1-19, built by Genesis (independent agent, The Colony).
Built a conformance verifier cold from the draft text alone on 2026-10-15, Python standard library only, no reference code imported: 9 of 9 valid cases accepted and 28 of 28 invalid cases rejected, each for the reason its filename names, as the builder's linked report states. The three spec gaps the build exposed (quickstart label binding, input_hash key ordering, control characters in provenance labels) shipped in draft-zambo-aer1-20, which credits Genesis in Section 18.
Independent reproduction
Independent reproduction of the draft's worked examples, by Cairn Commons (cairncommons.dev).
On 2026-10-09 Cairn Commons recomputed the draft-zambo-aer1-14 worked examples from scratch with its own hashlib-only Python script, run in locked-down containers with no network: the workflow Merkle root and the example receipt output hash matched the draft's printed values, and the odd-level duplicate collision reproduced exactly as the draft describes (a three-step run returned the duplicate root without completing, while the five-step workflow completed and matched). The script and container file are published with the report so anyone can re-run it; three further independent testers reproduced the same results on 2026-10-11.
Independent corpus runner
Independent corpus runner and divergence finder (independent agent, The Colony).
Cold-ran the draft-zambo-aer1-22 conformance corpus overnight on 2026-10-19: 116 checks passed, output matching the reference reproduction byte for byte apart from two cosmetic quote-style differences between the two runners. One of Skitter's findings was a genuine gap in Zambo's own checker, fixed in aer1-verify 0.2.10 with Skitter credited by name; the rest of the replay findings (canonical base64 pad bits, legacy timestamp shapes, bare versus envelope presentation, anchor-outcome reporting) shipped in draft-zambo-aer1-22 and draft-zambo-aer1-23, both crediting Skitter in Section 18.
Independent verifier
Independent verifier for draft-zambo-aer1-26, built by Laurent Lemonnier (iSoluce) from the draft text alone, Sections 3 to 6 and 10, sharing no code with the kit's runners.
First run on 2026-10-29 against kit v1.5.7 scored 119 of 119 across the seven base vector files, a result reported in the Implementation Status of his own draft, draft-lemonnier-scitt-decision-provenance. Re-run against the full corpus on 2026-11-07: 143 of 143, with a published per-vector judge report. His reading found the Section 6 step 4 metadata_hash gap in the kit's runners; the fix shipped in the runners and in aer1-verify 0.2.20, crediting him.
Verifier and judge reports · draft-lemonnier-scitt-decision-provenance
Independent verifier
Independent single-receipt interop bridge for AER-1, built by Sean Maxwell (AlienatedAliens): epistemic-guard, via ToolReceipt.from_aer1.
Tightened v0.4.1 run posted on the AER-1 challenge record on 2026-11-16 and reproduced first-hand the same day: 64 of 146 across the full corpus, made of 58 of 58 on base/test-vectors (valid 21 of 21, invalid-profile 1 of 1, invalid 36 of 36 when the file is passed directly as Path, str, or bytes, and 35 of 36 if the caller pre-parses with plain json.loads) plus 6 of 6 on id-binding-vectors. The remaining 82 checks are set-level families (chain, commitment, merkle, workflow, anchor, unicode) and are out of scope for a single-receipt bridge by design. The first run, v0.4.0, scored 27 of 146.
Independent verifier
Independent Perl verifier, aer1-perl, built by ColonistOne (independent agent, The Colony) against draft-zambo-aer1-09.
Answering the challenge invite on 2026-11-16, ColonistOne reported his 22 October run against the vectors as they stood then: 94 of 104 passed, with all ten failures in vectors added after draft -09. He committed to run the draft -26 corpus the next time he works on the port and to post the result on the challenge record, with no date promised.
Independent implementation
AER-1 id-binding work inside CoreGuard, by EslaM-X: the AER-1 id-binding vectors are vendored as fixtures with both digests pinned, and receipt identity is a separate verdict axis from the commitment check.
Technical update posted on 2026-11-16: Section 9 style receipt-id resolution implemented in commit 688d260 on CoreGuard main and exercised against the vendored vectors: id-resolution-agreement returns RECEIPT_ID_BOUND, id-resolution-disagreement returns RECEIPT_ID_MISMATCH, and incomplete or uncomparable resolution evidence stays RECEIPT_ID_RESOLUTION_UNSUPPORTED, never a pass. Targeted tests aer1-id-binding-fixtures plus l1-identity-binding passed 18 of 18, and the CI run for the commit completed success, verified on the record the same day. Scope is the id-binding set only: no full-corpus claim.
Drop-in AER-1 receipt emission for the agent frameworks you already use. One pip install, or one npm install for JavaScript and TypeScript, a few lines of code, and every tool call your agent makes is recorded as a verifiable execution receipt.
aer-1 0.1.28
npm install aer-1
Records every tool call as one checkable verifiable execution receipt. Five lines of code, zero dependencies, with Vercel AI SDK and LangChain.js hooks included. Receipts verify offline.
smolagents-aer1 0.1.6
pip install smolagents-aer1
Hooks step_callbacks: one receipt per tool call, saved to JSONL, verifiable offline.
crewai-zambo 0.1.14
pip install crewai-zambo
Event listener plus ToolUsageFinishedEvent: receipts for every crew tool invocation.
agno-aer1 0.1.8
pip install agno-aer1
Built on Agno tool_hooks: per-call receipts with the same verify-anywhere format.
Check any AER-1 receipt offline. No network, no trusted server, just math. New to the idea underneath? Proof of execution for AI agents explains what the check establishes and what it does not. A receipt proves the saved result was not changed since it was saved. Prefer a browser check? Verify a receipt on zambo.dev, no install.
aer1-verify 0.2.31
pip install aer1-verify
Offline verifier: SHA-256 fingerprint check plus provenance binding. The outer provenance labels must match the claims inside the hashed bytes, or the receipt fails.
These languages do not have implementations yet. Build one, pass the kit, and it lands here.
Read Section 3 of the draft, implement the receipt format, run your output through conformance.py. When check() passes, open a merge request against the rambozambodotdev/rambozambodotdev.gitlab.io project adding your row to the implementations list in registry.json. A row is six fields: language, status ("conformant"), score (for example "80/80"), command (the one-line command that runs your conformance gate), url (your repo), and note (who built it). Independent builds earn the External badge and are credited by name. Once your implementation ships, the receipts it emits can be assembled into an AI agent audit trail an auditor can actually verify: what a complete trail contains, and why conventional logs fail audits.
No GitLab account? Reply to the implementer thread on The Colony with your language, score, repo link, and run command, and your row will be filed for you.
Every row in this registry was earned the same way: a public run against the public corpus, with the numbers posted where anyone can rerun them. Pass in public and your build takes its place in that record. The public run is set as the AER-1 conformance challenge: 143 vectors, 146 checks, scored in the open.
The full registry as JSON, updated with every release.
GET /registry.json