Token导航 LogoToken导航TokenDH.com
研究检索external-serviceclawhub未标认证来源可访问clear审计提醒

pctxpctx 搜索

Agent Skill

pctx 用于查找、检索和筛选相关信息,适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

2,519

周安装

106

GitHub Stars

公开资料未说明

下载量

882
OpenClaw

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

MIT-0

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:pctx(pctx 搜索)
来源仓库:https://github.com/cianbyrne1010/pctx
安装命令:
openclaw skills install pctx
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install pctx

简介

pctx 用于在 OpenClaw 中高效检索和筛选信息,支持关键词匹配与来源线索定位。

  • 适用于需要快速获取候选结果的研究、数据查找或任务规划场景。
  • 通过 clawhub 安装后,结合具体 README 使用其 API 或命令完成信息聚合。
  • 使用前需确认权限范围、维护状态及是否涉及联网、文件读写等操作。
  • 建议核对原始仓库文档以了解实际功能边界和使用限制。

SKILL.md

name
pctx
description
MCP aggregation and Code Mode execution layer for token-efficient agent workflows. Wraps portofcontext/pctx — connects agents to Linear, GitHub, and other MCP servers via a single local endpoint. Use when agents need to call MCP tools or batch multiple tool calls into a single TypeScript execution to reduce token cost.
homepage
https://github.com/portofcontext/pctx
metadata
{"clawdis":{"emoji":"🔌","requires":{"tools":["brew","npm","node"]}}}

pctx — MCP Aggregation & Code Mode Execution

What is pctx?

pctx is a local server that:

  1. Aggregates MCP servers — connects to Linear, GitHub, and other MCP backends behind one endpoint
  2. Code Mode — instead of sequential tool calls, agents write TypeScript that runs in a Deno sandbox; only the result comes back (up to 98% token reduction on complex workflows)

Live endpoint: http://127.0.0.1:8080/mcp Connected MCPs: Linear (42 tools), GitHub (41 tools) Config: ~/.config/pctx/pctx.json Logs: /tmp/pctx.log | /tmp/pctx.err


When to use

  • Agent needs to call Linear or GitHub tools via MCP
  • Agent is doing multi-step tool workflows (pctx batches them into one TypeScript call)
  • Any workflow where sequential tool calls return large intermediate payloads

Quick Start

# Check if pctx is running
{baseDir}/pctx-skill.sh status

# List connected MCP servers + tool counts
{baseDir}/pctx-skill.sh mcp-list

# Test a tool call
{baseDir}/pctx-skill.sh test linear linear_getOrganization
{baseDir}/pctx-skill.sh test github list_issues

Commands

Daemon Management

# Status — shows port, uptime
{baseDir}/pctx-skill.sh status

# Start daemon (launchd)
{baseDir}/pctx-skill.sh start

# Stop daemon
{baseDir}/pctx-skill.sh stop

# Restart
{baseDir}/pctx-skill.sh restart

MCP Server Management

# List all connected MCPs + tool counts + connection health
{baseDir}/pctx-skill.sh mcp-list

# Add an upstream MCP — examples:
# stdio MCP with npm package
{baseDir}/pctx-skill.sh mcp-add memory --command "npx" --arg "-y" --arg "@modelcontextprotocol/server-memory"

# stdio MCP with installed binary
{baseDir}/pctx-skill.sh mcp-add linear --command "mcp-linear" --env "LINEAR_API_TOKEN=your_token"

# HTTP MCP
{baseDir}/pctx-skill.sh mcp-add stripe https://mcp.stripe.com

# Remove an MCP
{baseDir}/pctx-skill.sh mcp-remove <name>

Config & Backup

# Backup pctx.json (always done automatically before mcp-add/remove)
{baseDir}/pctx-skill.sh config-backup

# Restore from backup (interactive if no timestamp given)
{baseDir}/pctx-skill.sh config-restore
{baseDir}/pctx-skill.sh config-restore 20260422-092733

Testing

# Test a specific MCP tool via pctx
{baseDir}/pctx-skill.sh test linear linear_getOrganization
{baseDir}/pctx-skill.sh test linear linear_getTeams
{baseDir}/pctx-skill.sh test github list_issues
{baseDir}/pctx-skill.sh test github get_file_contents

Install (idempotent)

# Run on new VMs or after full rollback
{baseDir}/install.sh

Calling pctx from Agent Code

The pctx server exposes an MCP endpoint at http://127.0.0.1:8080/mcp. Agents can call tools directly via JSON-RPC, or use Code Mode for batched TypeScript execution.

Simple tool call via curl

# Initialize + get session context
curl -s -X POST http://127.0.0.1:8080/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"my-agent","version":"1.0"}}}'

# Call a tool (Code Mode — batched TypeScript)
curl -s -X POST http://127.0.0.1:8080/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: text/event-stream" \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"execute_typescript","arguments":{"script":"const org = await linear.linear_getOrganization({}); const teams = await linear.linear_getTeams({}); return { org, teams };"}}}'

Code Mode — batch multiple calls

The execute_typescript tool is the key feature. Write one script wrapped in async function run() that calls all tools needed:

Important: field name is code (not script). Code must be wrapped in async function run() { ... }.

Namespaces: Linear (capital L), Github (capital G, no 'b').

// BAD — multiple round-trips (traditional MCP)
const org = await Linear.linearGetOrganization({});
// ... agent receives result, burns tokens ...
const teams = await Linear.linearGetTeams({});
// ... agent receives result, burns tokens ...

// GOOD — one Code Mode call, one result
async function run() {
  const org = await Linear.linearGetOrganization({});
  const teams = await Linear.linearGetTeams({});
  const issues = await Linear.linearSearchIssues({ query: "urgent" });
  return JSON.stringify({ org, teams, issues }, null, 2);
}

// GitHub example
async function run() {
  const branches = await Github.listBranches({ owner: "MJM-Agents", repo: "rolling-reno-theme" });
  return JSON.stringify(branches, null, 2);
}

Currently Connected MCPs

NameBinaryTransportAuthTools
linearmcp-linearstdioLINEAR_API_TOKEN42
githubgithub-mcp-server stdiostdioGITHUB_PERSONAL_ACCESS_TOKEN41

Rollback

Full rollback instructions: ROLLBACK-MCP-PCTX.md in workspace root.

Quick rollback — remove one MCP:

{baseDir}/pctx-skill.sh mcp-remove linear   # removes Linear, leaves GitHub
{baseDir}/pctx-skill.sh mcp-remove github   # removes GitHub, leaves Linear

Full nuclear rollback:

{baseDir}/pctx-skill.sh stop
brew uninstall pctx github-mcp-server
npm uninstall -g @tacticlaunch/mcp-linear
rm -rf ~/.config/pctx/ ~/Library/LaunchAgents/ai.openclaw.pctx.plist

Config

VariableDefaultDescription
PCTX_CONFIG~/.config/pctx/pctx.jsonPath to pctx config
PCTX_PORT8080Port pctx listens on
PCTX_HOST127.0.0.1Host pctx binds to
PCTX_BINauto-detectedPath to pctx binary

Notes

  • pctx.json is chmod 600 — contains API keys, never commit
  • Deno sandbox: 10s execution timeout, no filesystem/env/system access
  • OAuth not yet supported in pctx (v0.7.1) — remote Linear/GitHub MCPs require OAuth; use stdio/local for now
  • Config auto-backed-up before every mcp-add / mcp-remove
  • PAT registered under mjm-dex GitHub account — swap if needed

适合场景

01

OpenClaw 用户查找和安装 Skill 时

02

用户想查找某类 Agent Skill 时

03

需要根据任务场景推荐可安装能力包时

04

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

补充不同宿主或平台的使用分布数据

能力 5

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

OpenClaw

96.54%
按下载量换算851

安全审计

VirusTotal

可疑

ClawScan

可疑

Static analysis

通过

权限和风险

external-service

该 Skill 可能调用第三方服务、云服务或外部模型 API,使用前需要确认账号、额度、数据发送范围和服务条款。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills