Your library, your reasoning,
on your terms.
AnythingLLM-AAMT is a corpus-and-RAG fabric for self-instrumentation — the fork where Wings reasoning, the AAMT MCP, and a stable Reasoning Provider API turn a chat shell into a sovereign cognition layer.
One fork. Two delivery modes. Run it on your hardware, or let us run it for you.
A corpus-and-RAG fabric for self-instrumentation.
Most LLM shells optimize for output. This one optimizes for reflection. Point it at your library — your books, your journals, your trace data — and ask it to reason with you, not at you.
The Wings reasoning integration is the differentiator. Small local models (Llama-3.1-8B class) become rigorous multi-step reasoners by validating every step against the AAMT mathematical foundations — and the trace is a first-class artifact you can query, replay, and sign.
Run it yourself. Or let us run it for you.
Same code. Same Wings reasoning. Same AAMT MCP. The only difference is whose metal it lives on — yours, or ours.
Sovereign
Run it yourself. Your weights, your data, your hardware.
- Air-gapped tarball — no DNS, no egress, no callbacks
- Your weights, your data, your hardware
- MIT-licensed substrate; AAMT/Wings/CTK load over a clean process boundary
- Single docker-compose for the reference deployment
- Free for self-hosters; pay only if you want our hardware
184 MB · signed · sha256 verified
Cloud
Run it for me. Same fork, same reasoning — we operate the metal.
- Hosted multi-tenant — zero local setup, works in any browser
- Same fork, same Wings reasoning, same AAMT MCP — we run the boxes
- JWT-bridged into the AsAManThinks platform identity (one login, one billing pool)
- Tier-gated: Admin · AgentStudio Developer · Practitioner+ · Lifetime
- Cloud egress audit log — see exactly what left your workspace, when
Admin · AgentStudio Developer · Practitioner+ · Lifetime
Same fork in both columns. The Sovereign tarball and the Cloud control plane are built from the same git SHA — switching modes is a deployment decision, not a re-platforming.
Four nodes. Three protocols. One reasoning fabric.
The fork is a corpus-and-RAG fabric for self-instrumentation. Wings turns it from a chat shell into a reasoning system. AAMT grounds the reasoning. NeterGaiiaM speaks the citation back to you.
A reasoning system, not a chat shell.
Same capabilities in Sovereign and Cloud. The differences live below the API.
Workspaces over your library
Your books, your PDFs, your private corpora — chunked, embedded, and ready to cite. Per-workspace key scoping; no cross-workspace data bleed.
Wings Reasoner over MCP
Wings exposes itself as an MCP server. The fork hosts it natively. Multi-step reasoning, AAMT-grounded, with structured Odu traces at every layer.
AAMT MCP
Workspace ops, file/git/scan/feature-map tools, and 78 AAMT instruments — all behind a single MCP surface. Reachable from the fork, the IDE, or the terminal.
Reasoning Provider API
Drop-in extension protocol. Any reasoner plugs in through one stable surface. The fork doesn't lock you to Wings — it makes Wings the easiest choice.
Reasoning Ledger API
Every Wings pass produces a signed, queryable, replayable artifact. Ask: 'what corpus did the agent consult before issuing this conclusion?' — and get a definitive answer.
NeterGaiiaM voice citation
Connect the fork to NeterGaiiaM and the agent will speak its citations — quoting passages from your own library back to you, with the page reference attached.
Workspace skills
Compose an agent in AgentStudio over your workspace data, deploy it as a skill — your team uses it from chat, voice, or the IDE without leaving their surface.
Cryptographic attestation
Every reasoning trace is signed with a workspace key. Tamper-proof audit log. Defense, healthcare, and legal teams can finally say yes to AI assistance.
Policy-bounded execution
CTK enforces OPA/Rego policies at every tool call. No tool call permitted outside policy. Provable governance, not promised governance.
Same fork. Different surface area.
Your library deserves a reasoner you trust.
Pick the side that fits the work. The fork is the same on both. The only question is whose hardware it lives on.