neo

分类: 工具与效率 | 上传者: 4ier | 下载: 0 | 版本: v1.0（最新）
Turn any web app into an AI-callable tool. Chrome extension passively captures APIs (fetch, XHR, WebSocket, SSE) with DOM trigger correlation. CLI to query, replay, and execute captured APIs with automatic authentication. Use when interacting with websites, discovering APIs, or replacing slow browser-use with direct API calls.
更新日志: Source: GitHub https://github.com/4ier/neo
目录结构

当前层级: skill/
📄 SKILL.md 7.3 KB
SKILL.md

---
name: neo
description: Turn any web app into an AI-callable tool. Chrome extension passively captures APIs (fetch, XHR, WebSocket, SSE) with DOM trigger correlation. CLI to query, replay, and execute captured APIs with automatic authentication. Use when interacting with websites, discovering APIs, or replacing slow browser-use with direct API calls.
---

# Neo — Web App API Discovery & Execution

## What
Chrome extension that passively captures web app APIs + CLI to query/execute them.
Captures fetch, XHR, WebSocket, and EventSource/SSE traffic with DOM trigger correlation.

## When to use
- Before browser-use: check if Neo has API knowledge for the domain
- User asks to interact with any website (GitHub, X, SaaS)
- User asks what APIs a website uses
- Replaying or debugging API calls

## CLI

Run from the repo root: `node tools/neo.cjs <command>`

Below, `neo` is shorthand for `node tools/neo.cjs` (add an alias or symlink for convenience).

```bash
# Overview
neo status                              # Domains + capture counts

# Capture management
neo capture summary                     # Quick overview (best for AI agents)
neo capture list [domain] [--limit N] [--since 1h]  # Recent captures (shows IDs)
neo capture search <query> [--method POST] [--status 200] [--limit N]
                                        # Search by URL pattern
neo capture count                       # Total count
neo capture domains                     # Domains with counts
neo capture detail <id>                 # Full capture details
neo capture stats <domain>              # Method/status/timing breakdown
neo capture watch [domain]              # Live tail (like tail -f)
neo capture clear [domain]              # Clear captures
neo capture export [domain] [--since 1h] [--format har]  # Export as JSON or HAR 1.2
neo capture import <file>               # Import from JSON (cross-device)
neo capture prune [--older-than 7d]     # Delete old captures
neo capture gc [domain] [--dry-run]     # Smart dedup: keep one per unique pattern

# Schema (API knowledge base)
neo schema generate <domain>            # Generate + save schema (with diff + versioning)
neo schema generate --all               # Batch generate for all domains with captures
neo schema show <domain>                # Human-readable with triggers
neo schema show <domain> --json         # Raw JSON
neo schema list                         # All known schemas (with versions)
neo schema search <query>               # Search all schemas by endpoint/path
neo schema coverage                     # Show schema vs capture coverage
neo schema openapi <domain>             # Export as OpenAPI 3.0 spec

# Execute
neo exec <url> [--method POST] [--header "K:V"] [--body "{}"] [--tab pattern] [--auto-headers]
                                        # Execute fetch in browser (auto auth)
neo api <domain> <search-term> [--body '{}']  # Smart API call: schema lookup + auto-auth + auto-tab
neo replay <id> [--tab pattern]         # Replay a captured API call

# Analyze
neo flows <domain> [--window <ms>] [--min-count <n>]
                                        # Discover API call sequence patterns
neo deps <domain> [--window <ms>] [--min-confidence <n>]
                                        # Discover API dependency chains (response→request data flow)
neo label <domain> [--dry-run]         # Add semantic labels to schema endpoints
neo workflow discover <domain>           # Discover reusable multi-step workflows
neo workflow show <name>                # Show workflow definition and transitions
neo workflow run <name> [--params k=v]  # Execute workflow step-by-step

# Page interaction
neo eval "<js>" --tab <pattern>         # Run JS in page context
neo open <url>                          # Open URL in Chrome
neo read <tab-pattern>                  # Extract readable page text

# Diagnostics
neo doctor                              # Check Chrome, extension, schemas, bridge
neo reload                              # Reload the Neo extension
neo tabs [filter]                       # List open Chrome tabs
neo mock <domain> [--port N] [--latency ms]  # Mock server from schema
neo suggest <domain>                    # Suggest AI capabilities for a domain
neo export-skill <domain>               # Generate agent-ready API reference (Markdown)
neo version                             # Show version
```

## Workflow: Find → Understand → Execute

```
1. neo capture search "CreateTweet"     # Find the API call
2. neo capture detail <id>              # Inspect full request/response
3. neo replay <id>                      # Replay it, or:
   neo api x.com CreateTweet --body '{}'  # Smart call (finds URL + auth automatically)
   neo exec <url> --auto-headers        # Execute with custom params
```

## Smart API Call (neo api)

```bash
neo api x.com badge_count               # GET badge count — zero config
neo api x.com HomeTimeline              # Finds endpoint, reconstructs URL, auto-auth
neo api github.com notifications        # Works across any domain with schema
neo api x.com CreateTweet --body '{"variables":{"tweet_text":"hello"}}'
```

`neo api` = schema lookup + capture URL reconstruction + auto-auth headers + auto tab selection.
Requires: captures exist + schema generated for the domain.

## Schema with Trigger Mapping & Field Variability

`neo schema show <domain>` shows which UI elements trigger which APIs and which body fields vary:

```
POST /i/api/graphql/:hash/CreateTweet  (12x, 340ms) [auth: x-csrf-token] body:{variables, features} [varies: variables]
  ← click button.tweet-btn "Post" (8x)
```

This maps: user intent → UI element → API call → parameterizable fields.

## Schema Knowledge Base

Schemas are saved alongside the CLI (e.g. `schemas/<domain>.json`) — local API knowledge.

Before calling an API, check schema: `neo schema show <domain>`
No schema? Generate: `neo schema generate <domain>` (requires captures)

Schemas are local-only, never in git. Personal browsing knowledge.

## Decision: Neo vs Browser-Use

1. `neo status` — does Neo have captures for this domain?
2. **Yes** → `neo exec` or `neo replay` (fast, precise)
3. **No** → `neo open <url>`, browse, captures accumulate, then use Neo
4. **Fallback** → browser-use for complex UI interactions

## WebSocket Bridge (real-time streaming)

```bash
neo bridge                              # Start bridge on ws://127.0.0.1:9234
neo bridge 9234 --json                  # NDJSON output (pipe to jq, etc.)
neo bridge --quiet                      # No stderr status messages
neo bridge --interactive                # Type commands to query extension
```

The extension auto-connects to the bridge. Features:
- **Real-time capture streaming**: see every API call as it happens
- **Bidirectional commands**: ping, status, capture.count, capture.list
- **NDJSON mode**: `neo bridge --json | jq .` for structured processing
- **Auto-reconnect**: extension reconnects if bridge restarts

Use bridge for long-running monitoring. Use CLI commands for one-off queries.

## Key facts
- Extension ID: `ikikhldfkbfmcbandaagjomhchlehjap`
- Extension dist: see repo `extension/` directory
- Source: https://github.com/4ier/neo
- API calls execute in browser context — cookies/auth inherited automatically
- Chrome must run with CDP on port 9222
- Per-domain capture cap: 500 entries, auto-cleanup
- Captures: fetch + XHR + WebSocket + EventSource/SSE
- Trigger tracking: click/submit → API call correlation (2s window)
登录后下载/点赞/收藏 ❤ 369 | ★ 0
neo

目录结构

SKILL.md

举报内容

提示
