M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
A

Agent Reach

Community
Give AI agents access to content across the internet.
How to install ↓ Compare GitHub source repository ↗
Category
Other #174 of 230
Stars
★ 99.1k Very popular
Runtime
Python 3.10+ · Node.js
Credentials
Optional API key
License
MIT
Last commit
46FMRS · D

The source presents Agent Reach primarily as a CLI-based multi-platform capability orchestration layer, not as a standalone MCP server with a published tool list and transport configuration. It suits users who want command-line agents to read and search internet content across many platforms; login-based platforms require careful account and credential risk assessment.

Strongest · Documentation 13/20 Weakest · Reliability 4/20

Reliability
4/20
Security and permissions
8/20
Maintenance
12/20
Documentation
13/20
Setup experience
9/20
Why each score
Reliability 4/20
The repository root provides no reviewable server manifest, and the README states that actual reading is delegated to multiple upstream CLI/MCP tools. Therefore MCP initialization, tool-list handshakes, and declared-versus-real behavior cannot be verified. The README claims backend probing and doctor diagnostics, but provides no committed tests, CI, or reproducible execution evidence; substantial points are deducted under the static-review rule.
Security and permissions 8/20
The README claims that cookies and tokens remain local with mode 600, and documents safe and dry-run modes as well as the account risk of treating cookies as full login credentials. Points are deducted because no implementation, data-flow evidence, permission boundary, or confirmation mechanism is reviewable. The default installation also auto-installs Node.js, gh CLI, mcporter, and other system dependencies and configures external services, so the claimed controls cannot be verified from the supplied material.
Maintenance 12/20
The project states that it uses the MIT license, is actively maintained, replaces failed backends, and provides an update path and GitHub Issues. Its non-archived status and large issue count indicate ongoing attention, but popularity is not scored. Points are deducted because the supplied material lacks verifiable release history, commit evidence, dependency-update policy, response targets, or a formal security-response channel.
Documentation 13/20
The README covers supported platforms, zero-configuration and login-state requirements, installation, updating, uninstall, safe mode, dry-run, diagnostics, cost notes, and capability limits, with several platform restrictions disclosed. Points are deducted because there is no manifest, tool schema, parameter/return-value documentation, error contract, or test evidence; many capabilities and backend selections remain unverified documentation claims.
Setup experience 9/20
The README provides a one-line installation flow aimed at Claude Code, OpenClaw, Cursor, and similar clients, and explains some permission, login-state, and server-proxy prerequisites. Points are deducted because the path depends on a remote installation document, runs pip and system-infrastructure installation, and provides no concrete MCP client configuration, connection example, or verified end-to-end evidence. Platforms requiring cookies or browser login state remain materially more complex.

Static review · not runListed 2026-07-29

Read the FMRS scoring method →

Fit and risk

What it can accessWrites / deletes local filesRuns commands or codeUses the network

Best for

  • Users who need Claude Code, OpenClaw, Cursor, or Windsurf to access content across multiple platforms
  • Individuals and developers seeking free or open-source upstream tools for web research
  • Agent workflows that need centralized installation, diagnostics, and backend routing

Not for

  • Tasks requiring graphical browser interaction, form submission, or automated login
  • Deployments requiring a standalone MCP server or a documented MCP transport
  • Environments that do not allow shell command execution by the agent
  • Use cases requiring every platform to work without login state or cookies

Required permissions

  • The agent must execute shell commands, including pip install, system dependency installation, and upstream CLI calls
  • Some channels require local browser login state, cookies, or tokens
  • Credentials are stored locally in ~/.agent-reach/config.yaml; the README states the file mode is 600

Risks and side effects

  • Cookie- or login-state-based access to Twitter/X, Xiaohongshu, Reddit, and similar platforms may trigger detection, restrictions, or account bans
  • The default installation may install Node.js, gh CLI, mcporter, and other system infrastructure, and modify agent skills and MCP configuration
  • Platform anti-crawling changes or upstream tool changes may break individual channels
  • Server deployments may require a proxy; the README estimates roughly $1 per month

Setup

Before you start

Runtime：Python 3.10+ · Node.js

TWITTER_AUTH_TOKEN requiredsecret Twitter/X auth token cookie, manually exported by the user via Cookie-Editor; must be explicitly set in the current shell before running the upstream `twitter` command.
TWITTER_CT0 requiredsecret Twitter/X ct0 cookie value, manually exported via Cookie-Editor and set alongside TWITTER_AUTH_TOKEN.

Give a command-capable agent the instruction “Install Agent Reach: https://raw.githubusercontent.com/Panniantong/agent-reach/main/docs/install.md”. Installation can use agent-reach install --env=auto; --safe enables safe mode and --dry-run previews operations. Run agent-reach doctor afterward to check channel status. OpenClaw users must enable exec permissions and restart the Gateway first.

Check that it works

After installation, run `agent-reach doctor`, which reports the status and active backend for each channel; alternatively ask the agent to fetch a YouTube video's subtitles to confirm connectivity.

Troubleshooting

  1. Run agent-reach doctor to inspect channel status, active backends, and repair advice
  2. OpenClaw users should verify tools.profile is coding and restart the Gateway
  3. Configure the required login state or cookies for platforms that need authentication
  4. Use agent-reach install --safe or --dry-run to inspect installation impact
  5. Check whether the affected channel has switched to a fallback backend

Things to try

Once connected, you can ask your AI assistant things like:

  • Summarize what this YouTube video says
  • Search Bilibili for AI tutorials
  • Read the content of this web page
  • Search the web for LLM framework comparisons

Use cases

Read arbitrary web pages and RSS/Atom feeds
Extract YouTube subtitles and search videos
Search and read public GitHub repositories
Search Twitter/X, Reddit, Xiaohongshu, and other platforms after configuring login state
Perform semantic web search through Exa

Supported clients

Claude CodePartial support
OpenClawPartial support
CursorPartial support
WindsurfPartial support

Listed from the project's documentation, not tested by this site.

Overview

Agent Reach is an open-source capability layer and CLI that selects, installs, checks, and routes upstream tools for web pages, YouTube, RSS, GitHub, Twitter/X, Bilibili, Reddit, Xiaohongshu, and other platforms. It is not an independent content-reading service: agents directly call the upstream tools for reading and search. The project requires Python 3.10+ and is licensed under MIT.

Similar servers

BioMCP 75 · B

One binary. One grammar. Evidence from the biomedical sources you already trust.

★ 659 · Tools 11 Compare with this →
Compare with other MCP servers

Source revision 94f06c1969df Data synced 2026-10-19 Read the FMRS scoring method