parallel-web-search Skill for Claude: What It Does and How to Install It
Makes Claude search the web with parallel-cli and answer with a link for every claim.
- Installs
- 14.6K
- on skills.sh
- Stars
- 91
- on GitHub
- Updated
- Oct 2026
- last repo commit
This is Parallel's own CLI-backed search skill. Claude turns your question into a parallel-cli search call, saves the full JSON to /tmp, and answers with inline links and a Sources list. SKILL.md says to prefer the Parallel Search MCP web_search tool for ordinary lookups when it is available.
It needs parallel-cli installed, signed in and funded on your Parallel account. The folder holds only SKILL.md and bundles no scripts.
Good for
- Anyone who wants cited web answers, with the raw results saved to a file
- Developers who want CLI-only search controls such as date and domain filters
- Users without the Parallel Search MCP tool who still want web search
Install parallel-web-search
Claude Code
- 1
Copy this prompt.
Install the parallel-web-search skill in Claude Code. Run this command: /plugin marketplace add parallel-web/parallel-agent-skills /plugin install parallel Then check that it loaded and tell me how to use it.
- 2
Paste it into Claude Code. It shows the change and asks before running it.
Pick one route. The plugin installs all 15 skills (counted from the repo), a hosted Parallel Search MCP server (parallel-search, anonymous) and a Bash-only agent called parallel-subagent. Restart Claude Code after installing it. This skill needs parallel-cli installed, signed in and funded; the second method in each tab lists the install commands from the repo's setup skill, and /parallel:parallel-cli-setup runs them for you. In Codex, use $ instead of / for skill names. This skill names the agent parallel:parallel-subagent in its frontmatter. That agent file (agents/parallel-subagent.md) ships with the plugin, not inside this skill folder, so the plugin is the route listed first.
From the project README. Source.
What's in the skill
When it loads
CLI-backed web search. Use when the user explicitly invokes this skill, needs CLI-only controls or saved JSON output, or no Parallel web_search MCP tool is available. When the bundled Parallel Search MCP is available, prefer its web_search tool for ordinary lookups and current-information queries. Only use parallel-deep-research if the user explicitly requests deep or exhaustive research.
Sections in SKILL.md
- Command
- Parsing results
- Response format
- Setup
Key facts
- Installs on skills.sh
- 14,605 (Nov 2, 2026)
- GitHub stars
- 91 for the whole repo (Nov 2, 2026)
- Last repo commit
- Oct 28, 2026
- Audits on skills.sh
- Gen Agent Trust Hub: pass, Socket: pass, Snyk: warn
- What it includes
- 1 file: SKILL.md. It bundles no scripts
What it includes
Runs one search from a plain-language objective, with optional keyword queries
Saves the full result to /tmp/<name>.json with -o and prints a trimmed view.
Try: “Search the web for the latest React 19 features and adoption”
Runs code ·
parallel-cli search--after-date, --include-domains and --exclude-domains narrow the results
Runs code ·
Date and domain filtersturbo for quick facts, fast for about one second, advanced for multi-step questions
Basic is the default. Fast needs parallel-cli 0.9.2 or newer, and latency is not guaranteed.
Runs code ·
Search modes--location takes a country code; --session-id groups related Search and Extract calls
Runs code ·
Location and sessionsEvery claim gets an inline link, and a Sources section ends the answer
SKILL.md says the Sources section is mandatory and links come only from the saved JSON.
Cited answer formatChecks exit status, warnings and empty results, and does not make up an answer
A stale output file is not evidence that a failed request worked.
Error handlingIf parallel-cli is missing, runs /parallel:parallel-cli-setup
Runs code ·
Setup fallback
From the official docs
Permissions
Medium risk
What it can do
- Read your data
- Create and change things
- Delete things
- Read-only mode
Watch out for
Your queries are sent to Parallel
The README says task inputs go to Parallel's API under its Customer Terms and Privacy Policy, and tells you not to send secrets or sensitive local content. Review what Claude puts in a request. github.com
It runs against your funded Parallel account
The README says CLI-backed skills are billed as ordinary API usage on your own Parallel account. Check current pricing on the vendor site before large runs. github.com
The setup path installs and signs in a CLI
The repo's setup skill installs parallel-cli (brew, uv, npm or pipx), starts a device sign-in, and points to `parallel-cli balance add` for top-ups. The search and research skills say never to add funds without explicit confirmation. github.com
Searches run in a Bash-only helper agent
The skill forks into parallel:parallel-subagent, whose file lists only the Bash tool and says it cannot read files. It runs the parallel-cli command and returns the output. github.com
Technical details
| Turning on read-only | There is no read-only switch. The skill runs parallel-cli commands that call Parallel's API under your account. |
|---|---|
| Sign-in and keys | parallel-cli signs in with a device OAuth flow you approve in a browser. The README says CLI-backed skills use credentials you already configured. |
| Scopes |
|
Questions
What does the parallel-web-search skill do?
It makes Claude run parallel-cli search and answer from the results, with a link for every claim and a Sources list at the end. The full results are saved to a JSON file for follow-up questions.
How do I install parallel-web-search?
Use the install box above. Install the Claude Code plugin first and restart Claude Code, then install and sign in to parallel-cli. The skill names an agent that ships in the plugin.
Do I need the CLI if I have the Parallel Search MCP server?
Usually not. SKILL.md says that when the bundled Parallel Search MCP is available, Claude should prefer its web_search tool for ordinary lookups. This skill is for when you invoke it yourself, need CLI-only controls or saved JSON output, or no web_search tool exists.
Is it free?
The bundled MCP search is free and anonymous within rate limits, per the README. This CLI skill runs against your own funded Parallel account and is billed as ordinary API usage. See the vendor's pricing page for rates.
parallel-web-search vs parallel-deep-research?
Search is the default for normal lookups. SKILL.md says to use parallel-deep-research only if you explicitly ask for deep or exhaustive research. The deep-research SKILL.md calls itself slower and more expensive.
Is it safe?
It runs parallel-cli in Bash and sends your queries to Parallel, which the README says you should keep free of secrets. It writes a JSON file under /tmp and spends credits on your account. Our advice: keep private details out of the search wording.
Does it work in Codex, Cursor or the Claude app?
The README documents Claude Code, Codex (with $skill-installer), Cursor, Cline and GitHub Copilot through parallel-cli skills install. It documents no Claude app upload, and the skill needs a terminal for parallel-cli.
Is it an official Parallel skill?
It is in Parallel Web Systems' own repo, and the repo's marketplace.json describes the plugin as the "Official Parallel Web Systems plugin for Claude Code".
Information
updated Nov 2, 2026
- Repository
- parallel-web/parallel-agent-skills
- Folder
skills/parallel-web-search
Sources
Docs, issue trackers and discussions this page is built from.
Show all 6Hide
- parallel-agent-skills README · github.com
- parallel-agent-skills marketplace.json · github.com
- parallel plugin.json · github.com
- parallel-cli-setup SKILL.md · github.com
- parallel-web-search SKILL.md · github.com
- parallel-subagent.md · github.com