https://github.com/deeefox/awesome-agent-search
Research and best practices for bringing web search into Codex, Claude Code, OpenClaw, and AI coding agents.
https://github.com/deeefox/awesome-agent-search
List: awesome-agent-search
agent-search agent-tools ai-agents awesome awesome-list brave-search claude-code codex coding-agents developer-tools llm-tools mcp model-context-protocol openclaw privacy retrieval search searxng tavily web-search
Last synced: 8 days ago
JSON representation
Research and best practices for bringing web search into Codex, Claude Code, OpenClaw, and AI coding agents.
- Host: GitHub
- URL: https://github.com/deeefox/awesome-agent-search
- Owner: DeeeFOX
- License: mit
- Created: 2026-05-21T17:45:21.000Z (29 days ago)
- Default Branch: main
- Last Pushed: 2026-05-22T14:20:35.000Z (28 days ago)
- Last Synced: 2026-05-22T14:24:29.972Z (28 days ago)
- Topics: agent-search, agent-tools, ai-agents, awesome, awesome-list, brave-search, claude-code, codex, coding-agents, developer-tools, llm-tools, mcp, model-context-protocol, openclaw, privacy, retrieval, search, searxng, tavily, web-search
- Language: JavaScript
- Homepage:
- Size: 223 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Contributing: CONTRIBUTING.md
- License: LICENSE
- Codeowners: .github/CODEOWNERS
- Security: SECURITY.md
- Roadmap: docs/roadmap/comparison-issues.md
- Agents: AGENTS.md
Awesome Lists containing this project
README
# Awesome Agent Search
English | [中文](README.zh-CN.md)
[](https://github.com/DeeeFOX/awesome-agent-search/actions/workflows/validate-registry.yml)
> Research and best practices for bringing external search into Codex, Claude Code, OpenClaw, and other coding agents.
A curated, public-safe, AI-readable research hub for agent search integration: native web search, MCP adapters, self-hosted search, hosted search APIs, browser/CDP retrieval, local documentation search, hybrid routing, evaluation, and implementation references.
## Contents
- [Why This Exists](#why-this-exists)
- [Search Options](#search-options)
- [For AI Agents](#for-ai-agents)
- [Why Awesome, Not Oh My](#why-awesome-not-oh-my)
- [Resources And References](#resources-and-references)
- [Appendix: Repository Navigation](#appendix-repository-navigation)
- [Contributing](#contributing)
## Why This Exists
Coding agents increasingly work against a changing external world: documentation, APIs, SDKs, release notes, package behavior, security advisories, product policies, issue threads, benchmarks, examples, and rendered web pages. Local repository context is necessary, but it ages quickly and cannot cover every public source an agent may need.
This repository helps answer a practical question:
> Which search route should an agent use for this task, and what are the tradeoffs?
It collects researched options, official entry points, best-practice notes, comparison criteria, and issue-ready research tasks so agent builders can choose between native search, MCP tools, self-hosted search, hosted APIs, browser retrieval, local documentation search, or hybrid routing without starting from a blank page.
## Search Options
This matrix links concrete options first. `Seeking` means this repository has not confirmed a durable install or deployment best-practice entry yet; `Researching` means a dedicated research report is still missing.
| Solution | Best Practice | Research Report | Strengths | Limitations | Agent Support Matrix |
| --- | --- | --- | --- | --- | --- |
| [OpenAI Web Search](https://developers.openai.com/api/docs/guides/tools-web-search) | Seeking | [Search integration strategies](docs/comparisons/search-integration-strategies.md) | Low setup when the selected OpenAI runtime exposes web search; integrated retrieval path. | Less operator control over backend choice, ranking, logs, routing, and reproducibility. | Codex |
| [Claude MCP Web Search](https://support.claude.com/en/articles/14503775-mcp-web-search) | Seeking | [Search integration strategies](docs/comparisons/search-integration-strategies.md) | Managed Claude-side route when the product or workspace supports it; minimal custom infrastructure. | Availability depends on Claude product, workspace, account, and region; backend policy is provider-managed. | Claude Code |
| [SearXNG](https://docs.searxng.org/) | [oh-my-agent-search](https://github.com/DeeeFOX/oh-my-agent-search) | [SearXNG search backend research](docs/research/searxng-search-backend-research.md) | Strong operator control over engines, JSON output, logs, retention, and deployment policy. | Requires operations; engine availability and result quality vary by environment; API-style access needs JSON enabled. | Claude Code Codex OpenClaw |
| [Brave Search API](https://brave.com/search/api/) | Seeking | [Brave Search API research](docs/research/brave-search-api.md) | Hosted independent-index web search with structured source URLs, freshness controls, country/language targeting, extra snippets, and reranking options. | Requires subscription-token management; cost/quota terms are volatile; privacy, retention, storage rights, and live result quality need review before primary use. | Codex Claude Code OpenClaw MCP agents |
| [Tavily API](https://docs.tavily.com/api-reference/introduction) | Seeking | Researching | Agent-oriented API shape; useful for structured retrieval and productized search workflows. | Hosted dependency; cost, quota, data policy, and answer abstraction can reduce transparency. | Codex Claude Code OpenClaw MCP agents |
| [DuckDuckGo Instant Answer API](https://duckduckgo.com/api) | Seeking | [DuckDuckGo Instant Answer API research](docs/research/duckduckgo-instant-answer-api.md) | No-key instant-answer JSON can support lightweight facts, entity lookups, definitions, and fallback routing signals. | Not a full web search results API; official documentation visibility is limited; language, freshness, coverage, and schema stability need stronger validation. | Codex Claude Code OpenClaw MCP agents |
| [eze-is/web-access](https://github.com/eze-is/web-access) | Seeking | [eze-is/web-access and Browser Retrieval](docs/research/web-access-browser-retrieval.md) | Handles rendered pages, navigation, interaction, screenshots, and page-state evidence. | Slower and more brittle than search APIs; browser profiles, cookies, sessions, and prompt injection need strict handling. | Claude Code |
## For AI Agents
This repository is designed to be read by agents, not only humans.
AI agents should:
1. Read [AGENTS.md](AGENTS.md).
2. Read [registry/resources.json](registry/resources.json).
3. Pick the closest guide, comparison, research note, or template.
4. Keep every query, endpoint, path, and example public-safe.
5. Keep runnable install commands in the installable companion, not in this awesome list.
6. Run `make review` before submitting changes.
Durable resources must be registered in `registry/resources.json`. Translations should use `language` and `translationOf` metadata.
## Why Awesome, Not Oh My
Use `awesome-agent-search` for this repository because it is a curated research and best-practices hub.
An `oh-my-*` name implies an installable package, plugin framework, or one-command enhancement bundle. Runnable setup belongs in companion repositories such as [oh-my-agent-search](https://github.com/DeeeFOX/oh-my-agent-search).
See [Companion repository boundary](docs/companion-repository.md) for the split between curated guidance and runnable setup.
## Resources And References
| Category | Resource | Use it for |
| --- | --- | --- |
| Core practice | [Best practices](docs/best-practices.md) | Query design, fallback, logging, citation, and safe contribution rules. |
| Strategy comparison | [Search integration strategies](docs/comparisons/search-integration-strategies.md) | Choosing between native search, MCP, SearXNG, hosted APIs, browser retrieval, local docs, and hybrid routing. |
| SearXNG research | [SearXNG search backend research](docs/research/searxng-search-backend-research.md) | Understanding when SearXNG fits agent search and which deployment pattern is appropriate. |
| Hosted API research | [Brave Search API research](docs/research/brave-search-api.md) | Understanding Brave Search API as a hosted independent-index search candidate for agents. |
| Fallback research | [DuckDuckGo Instant Answer API research](docs/research/duckduckgo-instant-answer-api.md) | Understanding DuckDuckGo Instant Answer API as a lightweight fallback, not a primary search backend. |
| Browser route | [eze-is/web-access note](docs/research/web-access-browser-retrieval.md) | Understanding browser/CDP retrieval as a representative Claude Code path. |
| MCP route | [Claude Code MCP guide](docs/integrations/claude-code-mcp.md) | Routing Claude Code search through explicit MCP tools. |
| Research queue | [Comparison issue backlog](docs/roadmap/comparison-issues.md) | Publishing focused comparison tasks as GitHub issues. |
| Registry | [Machine-readable registry](registry/resources.json) and [registry index](docs/registry-index.md) | Letting humans and agents discover durable resources. |
| Chinese docs | [中文 README](README.zh-CN.md) and [中文文档索引](docs/zh-CN/README.md) | Reading and searching the repository in Chinese. |
| Reference model | [sindresorhus/awesome](https://github.com/sindresorhus/awesome) | Curated-list conventions. |
| Reference model | [ohmyzsh/ohmyzsh](https://github.com/ohmyzsh/ohmyzsh) | Community docs and onboarding structure. |
| Reference model | [subinium/awesome-claude-code](https://github.com/subinium/awesome-claude-code) | AI coding-agent ecosystem taxonomy. |
| Reference model | [searxng/searx-instances](https://github.com/searxng/searx-instances) | Structured submission and review workflow. |
## Appendix: Repository Navigation
| Audience | Entry point | Use it for |
| --- | --- | --- |
| Humans | [Best practices](docs/best-practices.md) | Learn core agent search practices. |
| AI agents | [AGENTS.md](AGENTS.md) | Follow editing, validation, and privacy rules. |
| Integrators | [Integration guides](docs/integrations/README.md) | Connect agents to search tools, backends, and adapters. |
| Researchers | [Research notes](docs/research/README.md) | Read and contribute public-safe findings. |
| Maintainers | [Submission guide](docs/submission-guide.md) | Review issues, PRs, and registry entries. |
Repository layout:
- [docs/](docs): positioning, architecture, comparisons, integrations, i18n, research, evaluation, roadmap, launch, and submission guidance.
- [examples/](examples): minimal integration patterns by ecosystem.
- [registry/resources.json](registry/resources.json): machine-readable catalog.
- [templates/](templates): reusable integration and research templates.
- [schemas/](schemas): schema for structured registry data.
## Contributing
Read [CONTRIBUTING.md](CONTRIBUTING.md), [SECURITY.md](SECURITY.md), and [AGENTS.md](AGENTS.md) before opening a PR.
Run:
```sh
make review
```
Substantial contributions should state the problem, approach, expected benefit, privacy considerations, and acceptance criteria.
## License
MIT. See [LICENSE](LICENSE).