agentweb-mcp 1.2.0 → 1.2.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (3) hide show
  1. package/README.md +77 -29
  2. package/dist/index.js +1 -1
  3. package/package.json +1 -1
package/README.md CHANGED
@@ -5,11 +5,11 @@
5
5
  [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
6
6
  [![MCP Compatible](https://img.shields.io/badge/MCP-Compatible-blue)](https://modelcontextprotocol.io)
7
7
  [![Docker](https://img.shields.io/badge/Docker-supported-2496ED?logo=docker&logoColor=white)](https://github.com/zerabic/agentweb-mcp/blob/main/Dockerfile)
8
- [![Free for AI agents](https://img.shields.io/badge/Free%20for%20AI%20agents-0%2C-blueviolet)](https://agentweb.live)
8
+ [![Free for AI agents](https://img.shields.io/badge/Free%20for%20AI%20agents-0%24-blueviolet)](https://agentweb.live)
9
9
 
10
- MCP (Model Context Protocol) server for [AgentWeb.live](https://agentweb.live) — a free, agent-first business directory API with **12M+ businesses across 231 countries**.
10
+ MCP (Model Context Protocol) server for [AgentWeb.live](https://agentweb.live) — **the agent-native registry**. 11M+ businesses across 233 countries. Free, structured, machine-readable. **The substrate of the agent web** — a registry where businesses publish themselves directly, with no website required, designed from the ground up to be read by AI agents instead of browsers.
11
11
 
12
- > 🆕 **v1.1.0** — adds two-way write tools so agents can contribute new businesses, enrich existing ones, and report bad data. AgentWeb is now a self-healing network, not a one-way feed.
12
+ > 🆕 **v1.2.0** — adds `agentweb_get_short` for ~80% token savings via the public substrate (`/v1/r/{id}/agent.json`, no API key needed) and `agentweb_capabilities` for tool self-discovery. `get_business` now uses the canonical `/v1/r/{id}` endpoint with sub-50ms global latency.
13
13
 
14
14
  ## ⚡ 60-second Quickstart
15
15
 
@@ -36,29 +36,39 @@ MCP (Model Context Protocol) server for [AgentWeb.live](https://agentweb.live)
36
36
  # 3. Restart your client and ask:
37
37
  # "Find a coffee shop near Tivoli Copenhagen with phone and opening hours"
38
38
  #
39
- # Done. Your agent now has access to 12M+ businesses.
39
+ # Done. Your agent now has access to 11M+ businesses.
40
40
  ```
41
41
 
42
42
  ## What It Does
43
43
 
44
- This MCP server gives any Claude Desktop / Cursor / Continue / LangChain / OpenAI-Agents-SDK / etc. AI agent six tools:
44
+ This MCP server gives any Claude Desktop / Cursor / Continue / LangChain / OpenAI-Agents-SDK / Cline / etc. AI agent **eight tools**:
45
45
 
46
46
  ### Read tools
47
- - **`search_businesses`** — text + geo search (lat/lng + radius, city, country, category) across 12M+ records
48
- - **`get_business`** — full details for any business by ID
49
- - **`agentweb_health`** — service status + live counts + community contribution stats
50
- - **`agentweb_leaderboard`** — top contributing AI agents to the network
51
47
 
52
- ### Write tools 🆕
53
- - **`contribute_business`** — add a new business *or* enrich an existing one with phone, email, hours, address, etc. The directory deduplicates automatically by name+location and phone — your agent doesn't have to think about it.
54
- - **`report_business`** flag a business as closed, wrong info, or spam. Three or more "closed" reports automatically lower the trust score so other agents see it as unreliable. **Self-healing data quality loop, powered by agents.**
48
+ | Tool | Purpose |
49
+ |---|---|
50
+ | `search_businesses` | Text + geo search across 11M+ records (lat/lng + radius, city, country, category) |
51
+ | `get_business` | Full details for any business by ID. Now uses the canonical `/v1/r/{id}` substrate endpoint — sub-50ms global. |
52
+ | `agentweb_get_short` 🆕 | Compact agent-native record (~700 bytes, no API key needed). **80% fewer tokens** vs `get_business` for the same business. |
53
+ | `agentweb_health` | Service status + live counts + community contribution stats |
54
+ | `agentweb_leaderboard` | Top contributing AI agents to the network |
55
+ | `agentweb_capabilities` 🆕 | Self-introspection: list every endpoint, every tool, every data source |
56
+
57
+ ### Write tools
58
+
59
+ | Tool | Purpose |
60
+ |---|---|
61
+ | `contribute_business` | Add a new business *or* enrich an existing one (auto-dedupes by name+location and phone) |
62
+ | `report_business` | Flag a business as closed, wrong info, or spam. 3+ closed reports auto-lower trust. **Self-healing data quality loop, powered by agents.** |
55
63
 
56
64
  ## Why use it
57
65
 
58
- - **Free for AI agents.** Free signup, no credit card. Generous rate limits for personal/dev use.
59
- - **Designed for LLMs.** Clean JSON, predictable schema, structured everywhere. No HTML scraping required.
60
- - **Open data foundation.** Built on [OpenStreetMap](https://www.openstreetmap.org/) (ODbL-licensed) plus direct schema.org JSON-LD scraping of business websites no Google Places dependency, no legal grey areas.
61
- - **Two-way network.** Your agents can both *read* and *write back*every contribution improves the directory for every other agent.
66
+ - **Free for AI agents.** Free signup, no credit card. Generous rate limits.
67
+ - **Public substrate.** `agentweb_get_short` reads need **no API key at all** sub-50ms global, 700 bytes per record.
68
+ - **Designed for LLMs.** Clean JSON, predictable schema, single-letter compact format available. No HTML scraping required.
69
+ - **Open data foundation.** Built on [OpenStreetMap](https://www.openstreetmap.org/) (ODbL) plus direct schema.org JSON-LD scraping of business websites no Google Places dependency, no legal grey areas.
70
+ - **Two-way network.** Your agents both *read* and *write back* — every contribution improves the directory for every other agent.
71
+ - **Owner-claimable.** Business owners can claim and edit their listing in 30 seconds with **no website required** — agents can help the long tail of small businesses get visible.
62
72
 
63
73
  ## Installation
64
74
 
@@ -103,7 +113,7 @@ Add to `claude_desktop_config.json`:
103
113
  }
104
114
  ```
105
115
 
106
- ### Cursor / Windsurf / Continue
116
+ ### Cursor / Windsurf / Continue / Cline
107
117
 
108
118
  Same config block — these clients all read the same MCP format:
109
119
 
@@ -125,7 +135,7 @@ Same config block — these clients all read the same MCP format:
125
135
 
126
136
  ### `search_businesses`
127
137
 
128
- Search across 12M+ businesses by text query, category, city, country, or geographic radius.
138
+ Search across 11M+ businesses by text query, category, city, country, or geographic radius.
129
139
 
130
140
  **Parameters** (all optional, but at least one of `q`, `category`, `city`, or `lat+lng+radius_km` is required):
131
141
  - `q` — text search query
@@ -140,16 +150,38 @@ Search across 12M+ businesses by text query, category, city, country, or geograp
140
150
 
141
151
  ### `get_business`
142
152
 
143
- Get full details for a business by its UUID.
153
+ Get full details for a business by its UUID. Uses the canonical `/v1/r/{id}` substrate endpoint (no API key, sub-50ms global) with fallback to the legacy `/v1/business/{id}`.
144
154
 
145
155
  **Parameters:**
146
156
  - `id` — business UUID returned by `search_businesses`
147
157
 
158
+ ### `agentweb_get_short` 🆕
159
+
160
+ Same business as `get_business` but in the **compact shorthand format**: ~700 bytes instead of 3-5 KB, single-letter keys (`n=name`, `p=phone`, `e=email`, `w=website`, `h=hours`, `g=[lat,lng]`, etc.). **80% fewer tokens** for the LLM. **No API key required** — public substrate.
161
+
162
+ **Use this whenever you just need contact info.** Example: "Find me 10 cafes near Berlin and tell me their phones" — use `search_businesses` to get the IDs, then `agentweb_get_short` to fetch each one. The token savings add up fast.
163
+
164
+ **Parameters:**
165
+ - `id` — business UUID
166
+
167
+ Schema is self-describing: see https://api.agentweb.live/v1/schema/short
168
+
148
169
  ### `agentweb_health`
149
170
 
150
171
  Service health, total business count, country count, and community contribution stats.
151
172
 
152
- ### `contribute_business` 🆕
173
+ ### `agentweb_capabilities` 🆕
174
+
175
+ Returns the machine-readable description of the entire AgentWeb API: every endpoint, every tool, every data source, current live counts. Use this when you need to discover what AgentWeb can do, or when you want to introspect the substrate. **No API key required.**
176
+
177
+ ### `agentweb_leaderboard`
178
+
179
+ Top contributing AI agents to AgentWeb. See which agents are most actively improving the network — proves the network is two-way.
180
+
181
+ **Parameters:**
182
+ - `limit` — number of contributors to return (default 10, max 50)
183
+
184
+ ### `contribute_business`
153
185
 
154
186
  Submit a new business or enrich an existing one. The directory auto-deduplicates by name+coordinates (within 100m) and by phone number, so your agent doesn't need to check first.
155
187
 
@@ -162,7 +194,7 @@ Submit a new business or enrich an existing one. The directory auto-deduplicates
162
194
 
163
195
  **Try:** *"This pizza place I just visited isn't in AgentWeb yet — add it for me"*
164
196
 
165
- ### `report_business` 🆕
197
+ ### `report_business`
166
198
 
167
199
  Flag a business as closed, having wrong info, or being spam. 3+ "closed" reports automatically lower the trust confidence score.
168
200
 
@@ -173,17 +205,11 @@ Flag a business as closed, having wrong info, or being spam. 3+ "closed" reports
173
205
 
174
206
  **Try:** *"That dentist closed last year — report it"*
175
207
 
176
- ### `agentweb_leaderboard` 🆕
177
-
178
- Top contributing AI agents to AgentWeb. See which agents are most actively improving the network.
179
-
180
- **Parameters:**
181
- - `limit` — number of contributors to return (default 10, max 50)
182
-
183
208
  ## Example agent workflows
184
209
 
185
- **Personal assistant agent:**
210
+ **Personal assistant agent (token-efficient):**
186
211
  > "Find me three coffee shops within walking distance of Tivoli that open before 8am."
212
+ > *Agent calls `search_businesses` → gets 3 IDs → calls `agentweb_get_short` 3× (no API key, 700 bytes each) instead of `get_business` (3-5 KB each). 80% token savings.*
187
213
 
188
214
  **Research agent:**
189
215
  > "Search for all dentists in Berlin Mitte, then for each one check the website and report any that 404 as `closed`."
@@ -191,6 +217,22 @@ Top contributing AI agents to AgentWeb. See which agents are most actively impro
191
217
  **Local business directory agent:**
192
218
  > "I scraped this site listing 200 restaurants in Lisbon. For each one, contribute it to AgentWeb with the phone and hours."
193
219
 
220
+ **Self-discovery agent:**
221
+ > "What can AgentWeb do?"
222
+ > *Agent calls `agentweb_capabilities`, gets back the full machine-readable API description.*
223
+
224
+ ## The substrate
225
+
226
+ AgentWeb is the **substrate of the agent web**. Read more in the [manifesto](https://agentweb.live/manifesto).
227
+
228
+ Beyond MCP, every business is also reachable as:
229
+
230
+ - **Public agent.json** at `https://agentweb.live/v1/r/{id}/agent.json` — no API key, sub-50ms, ~700 bytes
231
+ - **Human profile page** at `https://agentweb.live/r/{id}` — server-rendered HTML with embedded JSON-LD for crawlers
232
+ - **Directory browse** at `https://agentweb.live/directory` — by category (15,000+) and country (233)
233
+
234
+ Owner-claimable: any business owner can claim their listing or add a brand-new one in 30 seconds at `https://agentweb.live/claim` and `https://agentweb.live/claim/new` with no website required.
235
+
194
236
  ## Data sources
195
237
 
196
238
  | Source | License | Refresh |
@@ -198,17 +240,23 @@ Top contributing AI agents to AgentWeb. See which agents are most actively impro
198
240
  | OpenStreetMap (Overpass API) | [ODbL](https://www.openstreetmap.org/copyright) | Weekly per country |
199
241
  | schema.org JSON-LD direct from business websites | As published by site owner | Continuous |
200
242
  | Agent contributions (`/v1/contribute`) | CC-BY 4.0 | Real-time |
243
+ | Owner claims (`/v1/claim/new`) | CC-BY 4.0 | Real-time |
201
244
 
202
245
  No Google Places, no Foursquare, no scraping of paid APIs.
203
246
 
204
247
  ## Discovery / well-known endpoints
205
248
 
249
+ - Manifesto: https://agentweb.live/manifesto
250
+ - Directory: https://agentweb.live/directory
206
251
  - OpenAPI spec: https://api.agentweb.live/openapi.json
207
252
  - API docs: https://api.agentweb.live/docs
208
253
  - Capabilities: https://api.agentweb.live/v1/capabilities
254
+ - Shorthand schema: https://api.agentweb.live/v1/schema/short
255
+ - Sitemap: https://agentweb.live/sitemap.xml
209
256
  - AI plugin manifest: https://agentweb.live/.well-known/ai-plugin.json
210
257
  - MCP manifest: https://agentweb.live/.well-known/mcp.json
211
258
  - AI directives: https://agentweb.live/ai.txt
259
+ - LLM-friendly summary: https://agentweb.live/llms.txt
212
260
 
213
261
  ## Development
214
262
 
package/dist/index.js CHANGED
@@ -10,7 +10,7 @@ class AgentWebServer {
10
10
  this.config = config;
11
11
  this.server = new Server({
12
12
  name: "agentweb-mcp",
13
- version: "1.2.0",
13
+ version: "1.2.1",
14
14
  }, {
15
15
  capabilities: {
16
16
  tools: {},
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "agentweb-mcp",
3
- "version": "1.2.0",
3
+ "version": "1.2.1",
4
4
  "mcpName": "io.github.zerabic/agentweb",
5
5
  "repository": {
6
6
  "type": "git",