# Taproot: AT Protocol MCP (remote · mcp.atproto.at)

Read public AT Protocol profiles, records, threads, backlinks and lexicons. No API key required.

- Trust score: 70/100 (medium)
- Change this week: +3
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-09-20

## Components

- remote · `mcp.atproto.at`: 70/100 (this document), [markdown](https://verifymcp.io/servers/com-atproto-mcp-taproot/mcp.md), [page](https://verifymcp.io/servers/com-atproto-mcp-taproot/mcp)

## Channel facts

- Endpoint: `https://mcp.atproto.at/mcp`
- Transports: `streamable-http`
- Auth: `none`
- Version: `1.0.0`

## Trust breakdown

How this component scores in each security and reliability category. Every signal is checked automatically against the live server, and we only credit what we can confirm. Scores are 0–100 per category. Scoring method: https://verifymcp.io/docs/scoring (what has changed: https://verifymcp.io/docs/scoring/changelog)

Scored 2026-09-20.

- **Endpoint Security**: 63/100
  - The endpoint's TLS certificate is valid, in date, and uses a strong key.
  - Authorisation not fully verified: no authorisation is required to call this server, and 27 tool(s) never declared a destructiveHint. The MCP spec treats an absent hint as destructive by default, so we cannot call this surface safe.
  - HTTPS is enforced; there's no plaintext access path.
  - The HSTS (Strict-Transport-Security) header is present.
  - DNSSEC check failed: this domain isn't protected by DNSSEC.
- **Transport & Reachability**: 100/100
  - Verified streamable-http transport via a live MCP handshake.
- **Schema Quality & AI Usability**: 83/100
  - 100% of prompts and resources have a non-trivial description (not blank, and not just the item's name).
  - AI-judged instruction clarity (excellent).
  - Context-footprint check failed: tool/resource definitions use about 4847 tokens (~179/item across 27 items; 27 tools + 0 resources), over budget; trim descriptions and params.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 23/100
  - Stability observed for 7 of 30 days with no destabilising changes; credit accrues until the full window elapses.
- **Tool Coverage**: 100/100
  - 100% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 100% of tool parameters carry a description.
  - Structured output schemas are declared (100% of tools); any adoption earns full credit.
- **Tool Safety**: 100/100
  - No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.
  - We read all 27 captured tool definition(s), and no name or description among them implies an irreversible operation.
  - An AI judge read all 28 captured unit(s) of tool text and found none that tries to manipulate the model reading it.
- **Capabilities**: 60/100
  - Spec-recency check failed: implements MCP spec 2025-06-18; the latest is 2026-07-28.

## Install

### How do I install the Taproot: AT Protocol MCP server?

Taproot: AT Protocol MCP is a hosted endpoint at https://mcp.atproto.at/mcp, so there is nothing to install locally. Ready-made configuration for Claude, Cursor, VS Code, Codex and 5 more is on this page, copied from each client's own documentation.

### Claude

```bash
claude mcp add --transport http com-atproto-mcp-taproot 'https://mcp.atproto.at/mcp'
```

### Cursor

```json
{
  "mcpServers": {
    "com-atproto-mcp-taproot": {
      "url": "https://mcp.atproto.at/mcp"
    }
  }
}
```

### VS Code

```json
{
  "servers": {
    "com-atproto-mcp-taproot": {
      "type": "http",
      "url": "https://mcp.atproto.at/mcp"
    }
  }
}
```

### Codex

```toml
[mcp_servers.com-atproto-mcp-taproot]
url = "https://mcp.atproto.at/mcp"
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "com-atproto-mcp-taproot": {
      "type": "remote",
      "url": "https://mcp.atproto.at/mcp",
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add com-atproto-mcp-taproot --url 'https://mcp.atproto.at/mcp' --transport streamable-http
```

### Hermes

```yaml
mcp_servers:
  com-atproto-mcp-taproot:
    url: "https://mcp.atproto.at/mcp"
```

### Netclaw

```json
{
  "McpServers": {
    "com-atproto-mcp-taproot": {
      "Transport": "http",
      "Url": "https://mcp.atproto.at/mcp"
    }
  }
}
```

### Vellum

```bash
assistant mcp add com-atproto-mcp-taproot -t streamable-http -u 'https://mcp.atproto.at/mcp'
```

### Other

```json
{
  "mcpServers": {
    "com-atproto-mcp-taproot": {
      "type": "http",
      "url": "https://mcp.atproto.at/mcp"
    }
  }
}
```

The mcpServers block is a cross-client convention. Remote transports vary, so check your client's docs.

## Changelog

Every change recorded for this component, newest first. Days that predate change tracking, or that we cannot explain, say so: "we were watching and nothing happened" and "we were not watching" are different claims.

### 2026-09-20 (score 70, +1)

No change was recorded against any check on this day. Stability & Change Management went from 20 to 23. That category is still filling its 30-day observation window: 6 days of observed history at the previous scan, 7 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-09-18 (score 69, +1)

No change was recorded against any check on this day. Stability & Change Management went from 13 to 17. That category is still filling its 30-day observation window: 4 days of observed history at the previous scan, 5 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-09-15 (score 68, +1)

No change was recorded against any check on this day. Stability & Change Management went from 3 to 7. That category is still filling its 30-day observation window: 1 days of observed history at the previous scan, 2 at this one. The score rises as the window fills, whether or not the server changes.

### 2026-09-14 (score 67, 0)

- [functional improvement] Stability: unverified → 0.03

### 2026-09-13 (score 67)

First indexed and scored.

## MCP tools (27)

### `resolve` (~198 tokens)

Resolve any AT Protocol identifier to JSON. Accepts an at:// URI (at://<did-or-handle>/<collection>/<rkey>), a bare handle (alice.bsky.social), or a DID (did:plc:...). Returns a profile, a collection page, or a single record depending on how specific the input is. PDS-agnostic: works for any AT Protocol PDS, not just Bluesky.

Input parameters:

- `cursor` (string): Pagination cursor (collections only).
- `include` (array): Optional enrichments: 'schema' resolves the collection's lexicon; 'backlinks' lists who referenced the target; 'profile' adds the repo owner's display name, bio, and avatar URL.
- `limit` (integer): Page size (1-100, default 50). A cursor in the response means more pages exist.
- `uri` (string, required): An at:// URI, a bare handle, or a DID.

### `get_profile` (~107 tokens)

Get an account's repository overview (handle, DID, PDS, and the list of collections it contains). Input is a handle or DID.

Input parameters:

- `actor` (string, required): A handle (alice.bsky.social) or DID (did:plc:...).
- `include` (array): Optional enrichments: 'schema' resolves the collection's lexicon; 'backlinks' lists who referenced the target; 'profile' adds the repo owner's display name, bio, and avatar URL.

### `get_record` (~108 tokens)

Fetch a single record by its full at:// URI (must include collection and rkey). Returns the record's value, CID, and resolved identity.

Input parameters:

- `include` (array): Optional enrichments: 'schema' resolves the collection's lexicon; 'backlinks' lists who referenced the target; 'profile' adds the repo owner's display name, bio, and avatar URL.
- `uri` (string, required): Full at:// URI: at://<did-or-handle>/<collection>/<rkey>.

### `list_records` (~113 tokens)

List records in a collection for an account, with pagination. Provide the repo (handle or DID) and the collection NSID (e.g. app.bsky.feed.post).

Input parameters:

- `collection` (string, required): Collection NSID, e.g. app.bsky.feed.post.
- `cursor` (string): Pagination cursor from a previous page.
- `limit` (integer): Page size (1-100, default 50). A cursor in the response means more pages exist.
- `repo` (string, required): A handle or DID.

### `get_backlinks` (~121 tokens)

Index of who/what references a target,the inbound likes, reposts, quotes, replies, and follows,as COUNTS per category, via the Constellation backlink index. Target is a record's at:// URI or an account DID. This returns aggregate counts only; each category also includes a `source` string. To list the actual records and their author DIDs for one category (e.g. WHO liked a post), call list_backlink_records with that `source`.

Input parameters:

- `target` (string, required): A record at:// URI or a DID.

### `list_backlink_records` (~213 tokens)

List the actual records (author DID + verified handle + at:// URI, newest first) that reference a target for ONE category,i.e. WHO liked, reposted, quoted, replied to, mentioned, or follows it. Workflow: call get_backlinks first to get a category `source`, then pass it here. To read WHAT a referencing record says (e.g. the text of a mention or reply), call get_record on its at:// URI. Author handles are resolved via identity resolution (verified), not search.

Input parameters:

- `cursor` (string): Pagination cursor returned by a previous call (to get the next page).
- `limit` (integer): Page size (1-25, default 25). Use the returned cursor for the next page.
- `source` (string, required): A category `source` from get_backlinks, e.g. app.bsky.feed.like:subject.uri.
- `target` (string, required): The referenced record at:// URI or account DID (same target passed to get_backlinks).

### `count_records` (~141 tokens)

Count the total records in an account's collection (e.g. how many likes a user has given). Scans server-side so you don't page manually. AT Protocol has no cheap exact count, so for very large collections this returns exact:false with a lower-bound count and a cursor,call again with that cursor and ADD the counts to keep going.

Input parameters:

- `collection` (string, required): Collection NSID, e.g. app.bsky.feed.like.
- `cursor` (string): Resume cursor from a prior capped (exact:false) count. The new count covers records AFTER it,add it to your previous total.
- `repo` (string, required): A handle or DID.

### `search_accounts` (~166 tokens)

Search for AT Protocol accounts by handle or display name (paginated, via the Bluesky AppView app.bsky.actor.searchActors, so accounts unknown to that index may be missing). Returns candidate accounts with DID + handle, to turn a name into a DID. Results are search CANDIDATES, not verified handle-to-DID mappings: call verify_handle before trusting that a handle really belongs to a DID (resolve/get_profile only confirm an account exists, not that the handle is authentic).

Input parameters:

- `cursor` (string): Pagination cursor from a previous call.
- `limit` (integer): Page size (1-100, default 50). A cursor in the response means more pages exist.
- `query` (string, required): A handle fragment or display name, e.g. "alice".

### `get_account_history` (~71 tokens)

An account's history from its PLC audit log: handle changes, PDS migrations, signing-key rotations, and creation time. did:plc accounts only (did:web has no PLC log). Input is a handle or DID.

Input parameters:

- `actor` (string, required): A handle or DID (did:plc).

### `search_lexicons` (~462 tokens)

Browse or search the registry of published AT Protocol lexicon schemas. The response's `total` field is the exact registry size and `matched` is the exact match count. Substring matches the NSID or description (query 'profile' → anything containing 'profile'); prefix the query with a DOT for a suffix match (query '.profile' → only NSIDs ENDING in .profile, i.e. profile record types). Filter by schema kind with `type` — 'give me every space lexicon' is type='space' with NO query, since an NSID does not imply its type (space.saebyeok.permissionSet is a permission set; app.kimbia.circle is a space). Types: the spec's primary types are record, query, procedure, subscription, and permission-set; a main def may also be a non-primary type (object, token, string); space comes from proposal 0016 (permissioned data), which is not in the ratified spec; defs-only marks a schema file with no main definition. Every response carries a `typeCounts` breakdown of the returned matches (it sums to `matched`), so one call with no arguments shows which kinds exist and how many; when you filter by type, `otherTypes` reports what the filter excluded, so a too-narrow or misspelled type still shows you the real options. Without a query or type, returns the total plus a sample. Results are paginated by `limit`; to see matches beyond the first page, call again with the `offset` printed at the end of the list. Use resolve_lexicon for a specific schema's full definition.

Input parameters:

- `limit` (integer): Page size (1-100, default 50). A cursor in the response means more pages exist.
- `offset` (integer): Skip this many matches (pagination). Use the offset value printed after a truncated list.
- `query` (string): Optional substring to filter by, e.g. "profile" or "app.bsky".
- `type`: Keep only these schema kinds, e.g. "space" for permissioned space declarations, or ["record","procedure"]. Case-insensitive. Combine with query to search within a kind.

### `resolve_lexicon` (~68 tokens)

Explain what a record type means by resolving its lexicon schema. Input is a collection NSID (e.g. app.bsky.feed.post). Returns the schema's description and definition.

Input parameters:

- `nsid` (string, required): A collection NSID, e.g. app.bsky.feed.post.

### `describe_pds` (~89 tokens)

Describe a Personal Data Server: its DID, whether registration requires an invite, the handle domains you can register there, and ToS/contact links. Works for any PDS, not just Bluesky. Input is the PDS origin (e.g. https://pds.example.com).

Input parameters:

- `pds` (string, required): A PDS origin URL, e.g. https://pds.example.com.

### `list_repos` (~252 tokens)

List the accounts (repositories) hosted on a PDS, via com.atproto.sync.listRepos: each repo's DID and active/takedown status, paginated. Answers 'who is hosted on this PDS?' / 'how many accounts does this PDS have?'. By default returns DIDs only (one request, large page — best for counting); pass enrich=true to also resolve handles (slower, smaller page). Input is the PDS origin (e.g. https://pds.example.com). Works for any PDS.

Input parameters:

- `cursor` (string): Pagination cursor from a previous call.
- `enrich` (boolean): Resolve each DID to its handle. Default false — returns DIDs only in ONE request (large page, the fast path for listing/counting). Set true only when you need handles (slower, smaller page).
- `limit` (integer): Page size (1-1000, default 1000). With enrich=true the effective page is capped near 20 (each DID is resolved to a handle). Use the returned cursor for the next page.
- `pds` (string, required): A PDS origin URL, e.g. https://pds.example.com.

### `get_repo_status` (~75 tokens)

Liveness/moderation status of an account's repository: whether it is active, its current revision, and any takedown/suspended/deactivated/deleted state. Answers 'has @alice been taken down?' which get_profile can't. Input is a handle or DID.

Input parameters:

- `repo` (string, required): A handle or DID.

### `get_latest_commit` (~88 tokens)

The current head commit of an account's repository (com.atproto.sync.getLatestCommit): its commit CID and revision. The cheapest way to fingerprint a repo or detect whether it changed since you last looked, without downloading data. Complements get_repo_status (which has the rev but not the commit CID). Input is a handle or DID.

Input parameters:

- `repo` (string, required): A handle or DID.

### `verify_handle` (~78 tokens)

Bidirectionally verify that a handle is really owned by its DID, via DNS TXT and HTTPS well-known. Returns whether each method passed. Use this to check for impersonation/spoofing,search results and resolve give candidates, this gives cryptographic binding.

Input parameters:

- `handle` (string, required): A handle, e.g. alice.bsky.social.

### `get_did_document` (~77 tokens)

Fetch the raw DID document for any DID or handle: all aliases (alsoKnownAs), every service entry (PDS, labeler, custom AppView), and verification methods. Broader than get_profile, and works for did:web accounts. Input is a handle or DID.

Input parameters:

- `actor` (string, required): A handle or DID.

### `get_app_link` (~76 tokens)

Turn an at:// URI (or handle/DID) into shareable web links,a Bluesky/Blacksky app URL to open in a browser, plus the canonical PDS getRecord URL. Use when the user wants a clickable/pasteable link.

Input parameters:

- `uri` (string, required): An at:// URI, handle, or DID.

### `get_pds_directory_info` (~91 tokens)

A cached health card for a PDS from pds.directory: version, DID, approximate user count, reachability, last-checked time, available domains, and which relays index it. Input is the PDS hostname (e.g. pds.example.com). Best-effort/cached data.

Input parameters:

- `hostname` (string, required): A PDS hostname, e.g. pds.example.com.

### `get_relay_directory_info` (~79 tokens)

A cached overview of a relay from firehose.directory: counts of PDSes it indexes (total/active/offline/banned) and total accounts. Input is the relay hostname (e.g. bsky.network). Best-effort/cached data.

Input parameters:

- `hostname` (string, required): A relay hostname, e.g. bsky.network.

### `query_labels` (~209 tokens)

What moderation labels a labeler has applied to a subject (com.atproto.label.queryLabels),e.g. 'has @alice or this post been labeled spam/nsfw/hidden by labeler X?'. Anonymous and service-agnostic: the labeler's own service endpoint is found from its DID document (no hardcoded aggregator). Subject is a DID/handle (account labels) or an at:// URI (record labels); labeler is the handle or DID of a labeler service (e.g. moderation.bsky.app).

Input parameters:

- `cursor` (string): Pagination cursor from a previous call.
- `labeler` (string, required): The labeler service to ask,a handle or DID, e.g. moderation.bsky.app.
- `limit` (integer): Page size (1-100, default 50). A cursor in the response means more pages exist.
- `subject` (string, required): What to check: a DID/handle (account-level labels) or an at:// URI (record-level).

### `list_repos_by_collection` (~301 tokens)

Network-wide: which accounts publish a given collection/lexicon,e.g. 'who has an app.bsky.feed.generator (a custom feed)?' or who uses a custom lexicon. Via a relay's com.atproto.sync.listReposByCollection (anonymous). By default returns DIDs only (one request, large page); pass enrich=true to also resolve handles (slower, smaller page). For just a total, use count_repos_by_collection. NOTE: relay-served (defaults to the public Bluesky relay; pass `relay` to query another), not the account's own PDS. Pairs with search_lexicons / resolve_lexicon.

Input parameters:

- `collection` (string, required): A collection NSID, e.g. app.bsky.feed.generator.
- `cursor` (string): Pagination cursor from a previous call.
- `enrich` (boolean): Resolve each DID to its handle. Default false — returns DIDs only in ONE request (large page, the fast path for listing/counting). Set true only when you need handles (slower, smaller page).
- `limit` (integer): Page size (1-1000, default 1000). With enrich=true the effective page is capped near 20 (each DID is resolved to a handle). Use the returned cursor for the next page.
- `relay` (string): Optional relay origin to query instead of the default, e.g. https://relay.example.com.

### `count_repos_by_collection` (~177 tokens)

Count how many accounts across the network publish a given collection/lexicon,e.g. 'how many accounts have an at.glean.subscription?'. Walks a relay's com.atproto.sync.listReposByCollection with NO per-account handle resolution, so it's cheap and returns an exact total for typical collections. For very large collections it returns exact:false with a cursor,call again with that cursor and ADD the counts (like count_records). Relay-served (defaults to the public Bluesky relay; pass `relay` for another).

Input parameters:

- `collection` (string, required): A collection NSID, e.g. app.bsky.feed.generator.
- `cursor` (string): Resume cursor from a previous exact:false response.
- `relay` (string): Optional relay origin to query instead of the default, e.g. https://relay.example.com.

### `list_blobs` (~160 tokens)

List the blobs (uploaded media/attachments,images, video, etc.) hosted in an account's repository, via com.atproto.sync.listBlobs. Returns each blob's CID and its canonical PDS getBlob URL (PDS-native, works for any PDS,not a CDN). Answers 'what media does this account host?'. Records reference blobs by CID; use this (or get_blob_info) to turn a CID into a fetchable URL. Input is a handle or DID.

Input parameters:

- `cursor` (string): Pagination cursor from a previous call.
- `limit` (integer): Page size (1-100, default 50). A cursor in the response means more pages exist.
- `repo` (string, required): A handle or DID.

### `get_blob_info` (~128 tokens)

Get metadata for a single blob: its canonical PDS getBlob URL plus a best-effort content-type and byte size (via a HEAD request). Returns the URL, never the bytes. Use to turn a blob ref (CID) found in a record into a downloadable link and learn what it is. Input is the owning repo (handle or DID) and the blob CID.

Input parameters:

- `cid` (string, required): The blob CID (e.g. bafkrei...), as found in a record's blob ref.
- `repo` (string, required): A handle or DID (the blob's owning repo).

### `list_labelers` (~118 tokens)

List the labeler (moderation) services known to the labeler relay,each labeler's DID, resolved handle, and service endpoint. The discovery step BEFORE query_labels, which needs a labeler DID/handle up front: use this to find moderation.bsky.app and any third-party/regional labelers, then pass one to query_labels.

Input parameters:

- `cursor` (string): Pagination cursor from a previous call.
- `limit` (integer): Page size (1-20, default 20). Use the returned cursor for the next page.

### `get_thread` (~194 tokens)

Reconstruct the conversation around a post: its ancestor chain up to the thread root (following reply.parent, root-first) PLUS a bounded descendant reply TREE,replies, replies-to-replies, and so on, each tagged with its depth and parent,walked breadth-first and capped for cost. NOTE: descendant expansion assumes the Bluesky reply model (app.bsky.feed.post with reply.parent); ancestor walking follows reply.parent generically, but the reply tree is specific to that lexicon and will be empty for record types that model replies differently. Answers 'show me this whole conversation / what is this replying to?'. If the tree is truncated, use list_backlink_records on a specific post's URI to expand it further. Input is a post's at:// URI.

Input parameters:

- `uri` (string, required): A post's full at:// URI: at://<did-or-handle>/app.bsky.feed.post/<rkey>.

## Diagnostics

Captured diagnostic sections: TLS, DNSSEC, Authorisation, Transports. The full working is on the page: https://verifymcp.io/servers/com-atproto-mcp-taproot/mcp#diagnostics

## Score history

- 2026-09-20: 70
- 2026-09-19: 69
- 2026-09-18: 69
- 2026-09-17: 68
- 2026-09-16: 68
- 2026-09-15: 68
- 2026-09-14: 67
- 2026-09-13: 67

## Common questions

### What is the Taproot: AT Protocol MCP server?

Taproot: AT Protocol MCP is listed in the public MCP registry as com.atproto-mcp/taproot. Read public AT Protocol profiles, records, threads, backlinks and lexicons. No API key required. This page covers its hosted endpoint (https://mcp.atproto.at/mcp).

### Is the Taproot: AT Protocol MCP server safe to use?

Taproot: AT Protocol MCP scores 70 out of 100 on VerifyMCP. That is a record of what we were able to check automatically, not an endorsement. The category breakdown on this page shows every signal behind the number, including the ones we could not confirm.

### What tools does the Taproot: AT Protocol MCP server expose?

Taproot: AT Protocol MCP exposes 27 tools: resolve, get_profile, get_record, list_records, get_backlinks, and 22 more. Their descriptions and schemas cost roughly 3,960 tokens of context every time the server is loaded.

### Does the Taproot: AT Protocol MCP server require authentication?

No. We connected to Taproot: AT Protocol MCP without credentials and it answered, so anything it exposes is reachable by anyone who knows the address.

### Is the Taproot: AT Protocol MCP server still maintained?

Taproot: AT Protocol MCP is still listed as active in the MCP registry. We last reached this channel on 20 September 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.

## Links

- Remote endpoint: https://mcp.atproto.at/mcp
- Website: https://atproto-mcp.com/
- Changelog RSS feed: https://verifymcp.io/servers/com-atproto-mcp-taproot/mcp.xml
- Changelog JSON feed: https://verifymcp.io/servers/com-atproto-mcp-taproot/mcp.json
- HTML version of this page: https://verifymcp.io/servers/com-atproto-mcp-taproot/mcp