agenttru.st

AllegatAI bronze

allegatai.com

“Evidence layer for legal work: a hypergraph of sources and propositions where reliance is decided by a guard, not by the model. External agents read briefings, verify drafted answers claim by claim, and propose candidate propositions that a human accepts.

https://allegatai.com/.well-known/agent-card.json its card
Checked 2d ago pushNotifications, streaming

we checked this    the operator says this

Community rating 0 0 up · 0 down — sign in to vote

Verified by agenttru.st

Everything here is a check agenttru.st performed itself. Assurance, protocol, hosting and freshness are in the card above and are not repeated.

Certificate
Issued by Google Trust Services domain-validated
Valid until 20 Jan 2027 as of our check on 12 Nov 2026. A wildcard certificate: its other hosts are invisible here, because CT logs the wildcard, not them.
Control of the hostname was checked; nothing about who operates it.
Key ECDSA P-256, signed ECDSA-SHA256, issued via WE1 under GTS Root R4 — classical, as every publicly trusted certificate is today.
Names on this certificate (2)

allegatai.com, *.allegatai.com.

Connection
TLS 1.3 · X25519MLKEM768 — post-quantum hybrid key exchange, agreed with our checker on 12 Nov 2026. What this means
DANE / TLSA
Not verified (TLSA query returned RCodeNameError)
Discovery
Well-known document
AI-facing documents
Publishes /llms.txt — “AllegatAI” a curated map of the site's content for language models
Fetched from this host during verification. Neither document is how this agent was discovered.
First seen
5 Nov 2026
View verification details
Assurance
bronze Bronze — agent card fetched over HTTPS with a valid certificate
Protocols
A2A verified by handshake or card fetch, not merely advertised
Hosted in
? Unknown · Cloudflare, Inc. (AS13335)
The address did not geolocate — usually anycast hosting, where one address answers from many places at once.
Last checked
2d ago

What this agent says it can do

Declared in the agent's own card. agenttru.st has not tested whether it completes any of these tasks — the operator of allegatai.com controls every word below.

list matters

List all matters with id, name, jurisdiction, and transition revision.

evidencemcpread

get matter overview

Matter cockpit: counts, relied/candidate propositions, positions with computed strength, open proof gaps, recent transitions.

evidencemcpread

get briefing

Start here. A briefing compiled from the matter as it stands now for one question: matching propositions labelled CURRENT (relied and support still stands), INVALID (relied but support no longer stands) or NOT PROOF, with their contradictors, sources (quarantined/superseded), basis and proof-gap neighbours, positions, supersessions, and whether it was truncated. `text` is a compact rendering for direct reading.

evidencemcpread

search propositions

Keyword-search propositions in a matter. Returns text, epistemic class, reliance, current support status (current/invalid/not_relied), and independent support count. Only `current` propositions are established; candidates are NOT facts.

evidencemcpread

get proposition

Full detail for one proposition: support sources with provenance roots, basis propositions, contradictors, work products citing it, and legal elements requiring it.

evidencemcpread

verify answer

Check an answer you wrote about this matter against it, claim by claim. Every sentence must cite proposition ids ([p-id]); a cited id must exist, a claim stated as fact must rest on a CURRENT proposition, and the cited text must carry what the claim says (its terms, numbers and dates). Returns the answer with failing claims removed, plus per-claim status: grounded, non_claim, uncited (kept, flagged), unknown_citation, not_proof, unsupported, misattributed (stripped), cross_language (kept, flagged: the cited propositions are written in a different script from the claim, so its terms could not be checked; its numbers and dates were). For a question that asks for judgement over the record (rank, compare, advise), pass analysis: true and state each proposition's support status beside its id ("[p-id] [CURRENT]"): the sentence then stands when every id it names exists and carries the status stated, and is marked analysis, not record restatement. Answers are verified up to 32,000 characters; the result reports answe

evidencemcpread

traverse hypergraph

Traverse the evidence hypergraph from any node id (source, proposition, actor, position, gap, work product) up to maxDepth hops, both directions. Returns the neighborhood subgraph with typed edges (supports, stated_by, derived_from, contradicts, requires, has_element, closed_by, cites, cites_authority, supersedes).

evidencemcpread

get graph slice

Pull structure on demand: the bounded evidence subgraph around proposition ids and/or one legal element. Seeds (at most 8) are expanded both directions up to depth hops (at most 3); the result holds at most maxNodes nodes (at most 60), seeds first, with typed edges, and says when it was truncated. Use it after the briefing to see what supports, contradicts or requires a proposition.

evidencemcpread

list sources

List evidence sources in a matter with provenance state (usable/quarantined/withdrawn), hashes, and derivative lineage. `sha256` is the hash of bytes this system holds. An external (vault-held) source carries its custodian's declared hash and locator under `external`; its `integrity` is declared_only until bytes are attached and match, so a declared hash is never verified bytes.

evidencemcpread

get source text

Get extracted text of one source, in windows of 8,000 characters. Pass offset to page through a long source; the result reports offset, totalChars and truncated, so you always know whether more remains and nothing is silently cut. Empty when extraction has not produced text (e.g. needs OCR). Photos carry origin "ai_vision" and aiDerived: a model read the image; treat it as a lead, never evidence.

evidencemcpread

get position

One legal position with computed strength and per-element support status.

evidencemcpread

list proof gaps

Material proof gaps with closure conditions and state (open/satisfied/abandoned).

evidencemcpread

get next proof actions

Value-of-information ranking: which proof to buy next, scored by weak legal elements unblocked, with inspectable rationale.

evidencemcpread

get transition log

The append-only transition log (audit trail) for a matter, most recent last.

evidencemcpread

propose proposition

Propose a candidate proposition linked to one source. Lands as CANDIDATE only: a human owner/counsel must accept it through the guard before anything may rely on it. For a statement, supply statedById with a matter actor id and knowledgeBasis so attribution survives to acceptance. An anchor’s excerpt is verified against the source’s held text: a span that misses its quote is corrected to the quote (the reply then also carries anchorCorrected:true), and a quote the text does not contain is refused. Returns {accepted:true, propositionId, reliance} or, when the guard refuses it, {accepted:false, reasons:[{code, propositionId?, field?, hint}]}: repair what each hint says and call again.

evidencemcpwrite

register external source

Register a source held in an external vault by its locator and, when the custodian recorded one, its declared SHA-256, without uploading bytes. It is recorded as declared (or locator_only when no hash exists), never as verified bytes; a human can attach the file later and, if a hash was declared, it must match. A locator-only source stays ingesting until its text or bytes are attached. Returns {accepted:true, sourceId, integrity} or {accepted:false, reasons:[{code, field?, hint}]}: repair what each hint says and call again.

evidencemcpwrite

attach external text

Attach text extracted outside this system to an external (vault-held) source so it can be searched, asked about and proposed from. The text's own SHA-256 is recorded; the original's bytes stay with the custodian and the source's integrity does not change (declared_only / locator_only stays so). Refused for uploaded sources and for sources whose bytes are held. A locator-only source still ingesting becomes usable. Text over the index limit is refused, never truncated.

evidencemcpwrite

propose identity link

Propose that two records of the same type (two people, two organizations, two places…) may be one entity. Lands UNRESOLVED only: it is a candidate to look at, never a decision, and it grants nothing: no access, no reliance, no merge of the two records. A name appearing in many sources is not corroboration; give exact passages quoted word for word from the source text (or a finding) where each record appears, and a basis that says what beyond the name points to one entity. A person (owner or counsel) decides it as the same, distinct or still unresolved; you cannot resolve a link. Returns {accepted:true, linkId, state:"unresolved"} or {accepted:false, reasons:[{code, field?, hint}]}: repair what each hint says and call again.

evidencemcpwrite

Technical agent card

Copied from the agent's card. The operator controls these values; agenttru.st has not verified them.

Provider
Algodyne Corporation — what this agent says about itself; other agents claiming the same provider are not thereby related
Protocol
a2a
Version
0.2.0
Auth schemes
apiToken
Card completeness
complete all eight fields required by a2a.proto v1.0
View all card details
Capabilities
pushNotifications streaming
Agent card
https://allegatai.com/.well-known/agent-card.json

Operate this agent and would rather not be listed? Request removal.