# Hashgrid DNA

> A search engine that learns from feedback, over embeddings: one tool per API route.

- name: ai.hashgrid/dna
- version: 1.1.0
- connection class: R1 (autonomous)
- trust: 83/100 flags: no-repository
- quality: 79/100 (new)
- owner: not claimed
- descriptor: https://api.protogrid.dev/v1/servers/ai.hashgrid%2Fdna
- tools: https://api.protogrid.dev/v1/servers/ai.hashgrid%2Fdna/tools

**An agent can connect on its own if it already holds `DNA_TOKEN`.** Remote endpoint behind a static API key or header. The registry never sees the value; the agent fills the placeholder.

## Connection (mcpServers)

```json
{
  "mcpServers": {
    "dna": {
      "type": "http",
      "url": "https://dna.hashgrid.ai/mcp",
      "headers": {
        "Authorization": "Bearer ${DNA_TOKEN}"
      }
    }
  }
}
```

Placeholders to fill: `${DNA_TOKEN}`.

## Trust

- hygiene: 80
- liveness: 100
- freshness: 100
- provenance: 55
- drivers: +dns-namespace +website +reachable +uptime-100% +updated-1d-ago -no-repository
- Derived from observable signals (official registry feed and our own credential-free probes); no code audit performed.

## Quality 79/100 (new)

### protocol: 88 (weight 25)

- pass `protocol.modern`: supports 2026-07-28 (server/discover)
- pass `protocol.stateless`: answers without a session
- pass `protocol.transport`: streamable HTTP
- warn `protocol.list_ttl`: tools/list declares ttlMs 0, so clients re-fetch it on every use

### authorization: n/a (weight 15)

- n/a `auth.prm`, `auth.as_metadata`, `auth.cimd`: no remote uses OAuth
- n/a `auth.secret_in_url`: no templated URL

### tool hygiene: 71 (weight 30)

- pass `tools.descriptions`: every tool has a description
- pass `tools.description_length`: descriptions are concise
- pass `tools.schemas`: every tool has an object input schema
- fail `tools.annotations`: no tool declares readOnlyHint or destructiveHint, so clients cannot tell safe calls from risky ones
- fail `tools.directory_hints`: no tool declares all four of readOnlyHint, destructiveHint, idempotentHint and openWorldHint; missing: create_node (readOnlyHint, destructiveHint, idempotentHint, openWorldHint); create_edge (readOnlyHint, destructiveHint, idempotentHint, openWorldHint); get_grid (readOnlyHint, destructiveHint, idempotentHint, openWorldHint); …
- pass `tools.token_cost`: about 1,109 tokens to load every tool
- pass `tools.api_dump`: tools are not a one-to-one API dump

### stability: n/a (weight 15)

- n/a `stability.changes`, `stability.rug_pull`: tool history recorded for 0 of the 7 days needed

### dependencies: n/a (weight 15)

- n/a `deps.known_vulns`, `deps.mcp_sdk_version`, `deps.resolvable`: no npm or PyPI package

- tools: 12, about 1,109 tokens to load them all
- tool set hash: 7ad248b66d25c3ede1163062ab65dbc6
- badge: [![protogrid quality](https://protogrid.dev/badge/ai.hashgrid/dna.svg)](https://protogrid.dev/servers/ai.hashgrid/dna)
- Checks run on what our credential-free, read-only probes observe; tools are never called and no code audit is performed.

### Tool changes

_No tool definition changes recorded._

## Remotes

- https://dna.hashgrid.ai/mcp (streamable-http, auth api_key ${DNA_TOKEN}, reachable true, uptime30d 1)

## Packages

_none_

## Tools (12)

- `create_edge`: POST /nodes/{nid}/edges, the search: from query node `nid`, the best node it has not been shown yet,
among those whose key starts with `prefix` ("" for any), stored as a new edge: `id` is the found node's id,
`score` is null until you score it (update_edge). Null when nothing is left. Call again for the next.
- `create_node`: POST /nodes: add a node. An item gets a `key`, a path unique in the grid (e.g. "/items/42"), so searches
can find it; a query gets none. `emb` is its embedding, 1 to 2048 floats from any model; use one model per
grid. Returns its id and key (not its emb, to keep replies short; get_node has it).
- `delete_edge`: DELETE /nodes/{nid}/edges/{eid}: forget a result, so it can be suggested again; its score no longer
trains.
- `delete_node`: DELETE /nodes/{nid}: delete a node and every edge from or to it; its scores no longer train.
- `get_edge`: GET /nodes/{nid}/edges/{eid}: one result of query `nid`, and its score (null if not scored).
- `get_grid`: GET /: the grid the API key opens, by name.
- `get_node`: GET /nodes/{nid}: a node, with its key (null if it has none) and its embedding `emb`.
- `list_edges`: GET /nodes/{nid}/edges: a query's results so far, each an edge with the found node's `id` and its
`score` (null if not scored), in id order. `limit` is 1 to 1024; for the next page, pass the last id as
`cursor`.
- `list_nodes`: GET /nodes: the grid's nodes, each with its id, key and emb, in id order. `limit` is 1 to 1024; for the
next page, pass the last id as `cursor`.
- `put_edge`: PUT /nodes/{nid}/edges/{eid}: score how well node `eid` fits query `nid`, from -1 (wrong) to 1
(perfect), creating the edge if DNA never suggested it, e.g. scores you already have. Only scores train
DNA; a new score replaces the last.
- `update_edge`: PATCH /nodes/{nid}/edges/{eid}: score a result of query `nid` that create_edge returned, from -1 (wrong)
to 1 (perfect). Only scores train DNA; a new score replaces the last.
- `update_node`: PATCH /nodes/{nid}: change a node. `key`: a new path, or null to remove it, so the node can no longer be
found; leave it out to keep it. `emb`: a new embedding; leave it out to keep it. Its scores keep training,
on the new embedding. Returns its id and key.