# BotClaw · Agent Observer

> Read-only agent radar, weekly editions, sourced guides, scenario choices and revision sync.

- name: tech.botclaw/agent-observer
- version: 1.2.0
- connection class: R0 (autonomous)
- trust: 82/100 flags: no-repository
- quality: 79/100 (good)
- owner: not claimed
- descriptor: https://api.protogrid.dev/v1/servers/tech.botclaw%2Fagent-observer
- tools: https://api.protogrid.dev/v1/servers/tech.botclaw%2Fagent-observer/tools

**An agent can connect right now, no human step.** Remote endpoint, no authentication, reachable on the last probe.

## Connection (mcpServers)

```json
{
  "mcpServers": {
    "agent-observer": {
      "type": "http",
      "url": "https://botclaw.tech/api/mcp"
    }
  }
}
```


## Trust

- hygiene: 80
- liveness: 99
- freshness: 100
- provenance: 55
- drivers: +dns-namespace +website +reachable +uptime-97% +updated-19d-ago -no-repository
- Derived from observable signals (official registry feed and our own credential-free probes); no code audit performed.

## Quality 79/100 (good)

### 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: 100 (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
- pass `tools.annotations`: 11 of 11 tools declare readOnlyHint or destructiveHint
- pass `tools.directory_hints`: every tool declares readOnlyHint, destructiveHint, idempotentHint and openWorldHint
- pass `tools.token_cost`: about 2,149 tokens to load every tool
- pass `tools.api_dump`: tools are not a one-to-one API dump

### stability: 25 (weight 15)

- warn `stability.changes`: 7 tool changes in 30 days
- fail `stability.rug_pull`: 1 tools kept their name but changed most of their description in 30 days; review before trusting them

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

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

- tools: 11, about 2,149 tokens to load them all
- tool set hash: d12aee3870f36d1367b34d1fb38bb79f
- tools last changed: 2026-10-05T14:38:55.555Z
- badge: [![protogrid quality](https://protogrid.dev/badge/tech.botclaw/agent-observer.svg)](https://protogrid.dev/servers/tech.botclaw/agent-observer)
- Checks run on what our credential-free, read-only probes observe; tools are never called and no code audit is performed.

### Tool changes

- 2026-10-05 `botclaw_get_workflows` changed (description, outputSchema) **meaning changed**
  - before: Find owned, reproducible workflow procedures. Each list entry has a stable ID; call get_item for steps, pinned versions and verification limits. The RSS example and OpenClaw staging-backup trial have execution evidence; memory behavior and upgrade regressions remain untested. Does not run workflows.
  - after: Find owned, reproducible workflow procedures. Each list entry includes its recorded verification kind, evidence URLs and limits, independent of content format. Call get_item for steps and pinned versions. A software run proves only its documented checks, not end-to-end agent success. Does not run workflows.
- 2026-10-02 `botclaw_get_workflows` changed (description)
  - before: Find owned, reproducible workflow procedures. Each list entry has a stable ID; call get_item for the steps and actual verification boundary. Only the Node RSS example currently has execution evidence. Does not run workflows.
  - after: Find owned, reproducible workflow procedures. Each list entry has a stable ID; call get_item for steps, pinned versions and verification limits. The RSS example and OpenClaw staging-backup trial have execution evidence; memory behavior and upgrade regressions remain untested. Does not run workflows.
- 2026-10-02 `botclaw_get_sources` changed (outputSchema)
- 2026-10-01 `botclaw_choose` added
  - Apply deterministic task, hosting, platform, budget and access gates to published dossiers. Each candidate has sourced reasons, explicit unknowns and hosting conflicts. No performance scores, account-access guarantee, inference, purchase or setup action. no-new-subscription does not establish free inference.
- 2026-10-01 `botclaw_get_observations` added
  - Read published bilingual version radar, weekly editions, question guides or product dossiers. Includes full bodies, source dates, review dates and weekly coverage cutoff. Editorial interpretation is not product execution. Use product to include records discussing that product, including multi-product guides.
- 2026-10-01 `botclaw_get_changes` changed (outputSchema)
- 2026-10-01 `botclaw_get_item` changed (outputSchema)

Full history: https://api.protogrid.dev/v1/servers/tech.botclaw%2Fagent-observer/changes

## Remotes

- https://botclaw.tech/api/mcp (streamable-http, auth none, reachable true, uptime30d 0.972973)

## Packages

_none_

## Tools (11)

- `botclaw_choose`: Apply deterministic task, hosting, platform, budget and access gates to published dossiers. Each candidate has sourced reasons, explicit unknowns and hosting conflicts. No performance scores, account-access guarantee, inference, purchase or setup action. no-new-subscription does not establish free inference.
- `botclaw_compare`: Compare source-backed capabilities and explicit unknowns for the seven products. Filters can shortlist documented self-managed options. This is official documentation, not a tested product ranking or current account-access guarantee.
- `botclaw_get_catalog`: List the seven tracked products, category and content-type IDs, and the public MCP's paging and language options. Use these IDs in other tools.
- `botclaw_get_changes`: Read a sequence-ordered change stream after a decimal cursor. Start at 0, persist nextCursor only after applying the page, and continue while hasMore. Upserts contain the current published projection; withdrawn events contain only ID/revision/date, item=null. Remove withdrawn records locally. Replaying a page is safe. This is not a historical full-text archive.
- `botclaw_get_daily`: Read public items dated on a real YYYY-MM-DD calendar day in Asia/Shanghai. Omit date for today. This is a date index, not an AI-generated daily report. reviewedAt is a source-review date, not a publication date. Returns 20 items per page.
- `botclaw_get_item`: Read one published record by stable ID, with body, applicable conditions, evidence type, revision and canonical citation URL. Missing or withdrawn records return item=null. Original-source excerpts are not independent execution evidence.
- `botclaw_get_latest`: Read the latest dated agent news, newest first. Defaults to news; choose content_type guide, experience or review for other reading. Filters are listed by botclaw_get_catalog. Results have 20 items per page; use hasMore and page to continue.
- `botclaw_get_observations`: Read published bilingual version radar, weekly editions, question guides or product dossiers. Includes full bodies, source dates, review dates and weekly coverage cutoff. Editorial interpretation is not product execution. Use product to include records discussing that product, including multi-product guides.
- `botclaw_get_sources`: List public source URLs, source origin (official or independent), collection state, last successful collection, and item counts. Manual sources are editorially reviewed; they are not automatic collectors.
- `botclaw_get_workflows`: Find owned, reproducible workflow procedures. Each list entry includes its recorded verification kind, evidence URLs and limits, independent of content format. Call get_item for steps and pinned versions. A software run proves only its documented checks, not end-to-end agent success. Does not run workflows.
- `botclaw_search`: Search published titles, summaries and owned workflow text using bounded bilingual keyword aliases (1–80 characters). Product, content type and category filters apply. Returns up to 20 records per page, not a generated answer or live-internet search. Use get_item to inspect conditions, evidence and citation URLs.