walid@portfolio:~/lab/free-agent-web-access$
cd../lab
10ideaAug 2026

The free agent web access setup

One open-source install that lets an agent read YouTube, GitHub, X and Reddit

Read and search access to the platforms an agent normally cannot touch — YouTube transcripts, GitHub, X posts, Reddit threads, any URL as clean Markdown — from one install rather than a stack of paid APIs. The X API alone runs about $215 a month at moderate use; this is MIT and free.

MCPClaude CodeResearchYouTubeRedditGitHubOpen sourceMITAgent-Reach on GitHub ↗
What this is

One install instead of a pile of APIs

Agent-Reach is an open-source Python tool that gives an agent a working path into the platforms where the information density is highest — and where every one of them puts up a different barrier. The X API is pay-per-use at roughly $215 a month for moderate use. Reddit 403s server IPs. Plenty of sites simply block bots.

It bundles a tool per platform, with a primary and a fallback backend for each, so when an access route dies it switches without you noticing. MIT licensed, Python 3.10+, and it works with anything that can run commands — Claude Code, Cursor, Windsurf. Cookies stay on your machine and are never uploaded.

The payoff is speed: competitor research, content ideas mined from real complaints, sourcing quotes, or pulling a transcript to repurpose — one prompt instead of an afternoon of copy-paste across five tabs.

Step 1

Install it

There is no repo to clone by hand. You paste the first line to your agent and it does the rest.

agent-reach.txt8 lines
# paste this line to your agent — it installs the rest itself
Install Agent Reach: https://raw.githubusercontent.com/Panniantong/agent-reach/main/docs/install.md

# already installed, and something stopped working
Update Agent Reach: https://raw.githubusercontent.com/Panniantong/agent-reach/main/docs/update.md

# what works, what doesn't, and how to fix it
agent-reach doctor
Step 2

What works out of the box, and what needs a login

The zero-config half is the half worth building a habit on. Everything else costs you a cookie or a logged-in browser session.

PlatformWhat you getSetup
WebAny URL turned into clean Markdownzero config
YouTubeSubtitles and search, across 1800+ video siteszero config
GitHubRead and search; gh auth unlocks issues and PRszero config
Web searchSearch with no API keyauto-configured
RSSAny RSS or Atom feedzero config
V2EXHot topics, threads and replieszero config
X / TwitterRead, search, timeline, articlesbrowser cookie
RedditSearch and read — no anonymous path existssession or cookie
Instagram, FacebookProfiles, posts, searchlogged-in Chrome
LinkedInPublic pages, profiles, job searchsetup step
Step 3

The research prompt

Three sources, one brief. This is the shape that makes the setup worth wiring up.

research.prompt8 lines
Using the connected data tools:
1. Pull the transcript of this YouTube video: [URL] and summarize the 5 key
   takeaways.
2. Search Reddit for what people actually complain about with [product/topic]
   and group the complaints into themes.
3. Find 3 recent X threads on [topic] and tell me the consensus and the
   contrarian take.
Give me one combined brief.
!
Where the free part stops

X has tightened access and moved to pay-per-use, and Reddit blocks anonymous endpoints outright — both need a cookie or a logged-in browser session, and neither is reliable unattended. YouTube, GitHub, RSS and plain web reading are the genuinely zero-config ones. Build the habit on those and treat the rest as a bonus.

When it does not pull

Four failures and their fixes

Before any of these, run agent-reach doctor — it reports what works, what does not, and what to do about it.

SymptomThe fix
A tool returns nothingThat platform is rate-limited or its backend rotated. Run doctor, then try another source.
The agent cannot see the toolsThe install did not finish, or the agent needs a restart to pick the tools up.
The transcript is emptyThe video has no captions. Ask for auto-captions, or pick another video.
Results feel shallowThe query was too loose. Tighten it and ask for themes rather than a dump.
i
Rule of thumb

Lean on the zero-config sources — web, YouTube, GitHub, RSS — and let the agent fall back to them whenever a cookie-gated platform is blocked.

Do this in the next 10 minutes

The smallest version that proves it

✓paste the install line to your agent✓run agent-reach doctor and read what is green✓run the research prompt on one video and one repo
← previous
The five skills to install before anything else
next →
Five skills that turn Claude Code into an animation studio
← all experiments