An independent research agent working on Gandhāran and early Indo-Aryan material.
It acts under its own cryptographic identity, and every answer it gives carries the grade of the evidence behind it.
5ChAGI reads scholarly sources — inscriptions, numismatics, archaeology, ancient DNA — and the published manifests of other agents. It is operator-supervised: grading is automatic, publishing is not, and a human decides what leaves the system.
It does not act on behalf of anonymous end users, and it represents nobody but itself. Its accountability runs through a signing key you can check, not through a company name you would have to trust.
Most retrieval systems return a passage and leave you to judge it. This one returns a claim and what the claim rests on, on a four-term ladder that never promotes itself.
| Grade | What it means |
|---|---|
| verified | Cross-validated contemporary primary evidence, or ancient DNA. |
| attested | A firm witness, but resting on a consensus chain. |
| contested | Disputed authenticity or dating. The dispute stays visible. |
| tradition | Later textual transmission, not epigraphy. |
Model output is interpretive until independently checked, and is never promoted to verified by automation. The full commitments, including what it refuses to do, are published and machine-readable.
Discovery documents are open and require no key. The MCP interface is access-gated; counterparties are issued individually revocable credentials.
Outbound requests are signed under Web Bot Auth (RFC 9421 HTTP Message
Signatures, Ed25519) with Signature-Agent: "https://5chagi.ai".
Fetch the directory above and verify against the keyid.
A request claiming to be 5ChAGI without a verifying signature is not from 5ChAGI. We would rather be blocked than impersonated — how to block it is documented too.
Formal admixture analysis of a consumer genotype file against published ancient DNA — proportions with intervals, comparative placement, and a written statement of what the analysis will not claim. Scope and pricing.
Operational or abuse concerns, or a request for access: [email protected]. If this agent behaved badly on your host, we want the report — say so and it stops.