onesearch: A CLI Research Tool for AI Agents with Unified Search and Evidence Collection Workflow

onesearch is an open-source, CLI-first research and evidence collection tool designed for AI Agents, scripts, and terminal users. It integrates search, source discovery, document lookup, page fetching, site mapping, crawling, and repo wiki workflows into a unified interface. Built with Go language, it addresses the pain points of AI Agents in accessing external information efficiently and reliably.

With the rapid development of large language models and AI Agents, efficient and reliable external information access has become critical. Existing search tools are human-focused, with inconsistent interfaces and return formats, making them hard to integrate into AI Agent workflows. onesearch was created to solve this: it uses a CLI-first design to encapsulate search and evidence collection capabilities into a unified command-line interface, suitable for AI Agents, automation scripts, and terminal-savvy developers.

onesearch integrates multiple information acquisition capabilities:

Search & Discovery

• General search: Supports mainstream search engines with structured results.
• Source discovery: Automatically identifies and extracts original sources for traceability.

Docs & Knowledge Acquisition

• Docs Lookup: Quickly locates technical docs, API references.
• Page Fetch: Crawls specified URLs and parses various content types.

Site & Content Crawling

• Site Mapping: Generates website structure and page lists.
• Crawling: Automated content collection with configurable strategies.

Repo Wiki Integration

• Supports README parsing, Wiki page fetching, and code structure analysis for better project context understanding by AI Agents.

Pure Go Implementation Advantages

• High performance: Go's concurrency model and compiled nature ensure efficient execution for batch tasks.
• Single binary deployment: Static compilation allows easy deployment without dependencies.
• Cross-platform: Runs on Linux, macOS, Windows.

CLI-first Design Philosophy

• Standardized output: All commands return structured data (JSON) for programmatic parsing.
• Pipe-friendly: Seamlessly integrates with other CLI tools via Unix pipes.
• Scriptable: Stable command interfaces and exit codes for automation.
• Config-driven: Manages API keys, proxies via config files or environment variables.

AI Agent Information Access

• RAG system enhancement: Serves as a retrieval component for LLM external knowledge.
• Tool call integration: Enables LLMs to主动 search via function calling.
• Evidence chain construction: Automatically records sources for answer traceability.

Developer Research Workflow

• Technical research: Collects relevant materials quickly.
• Competitor analysis: Batch fetches product info and docs.
• Code review assistance: Pulls README, docs, and code structure.

Automation & Batch Processing

• Monitoring & intelligence collection: Regularly checks info source changes.
• Content aggregation: Collects from multiple sources to generate reports.
• Data preprocessing: Gathers training data for ML pipelines.

Feature	Traditional Search Tools	onesearch
Target Users	Human users	AI Agents + Developers
Interface Design	Interactive UI	CLI + Structured Output
Integration Capability	Limited	Native support for scripts/program calls
Function Range	Single search	Search + Fetch + Crawl + Repo Integration
Deployment	Complex dependencies	Single binary

1. API Key Management: Use environment variables or config files to securely manage API keys for search/crawling.
2. Rate Limits: Comply with target site robots.txt and rate limits to avoid being blocked.
3. Output Processing: Use tools like jq to handle JSON outputs.
4. Error Handling: Properly handle exit codes and error outputs in scripts for robust automation.

Project Status

onesearch is in early development (0 stars on GitHub) but aligns with AI Agent toolchain trends.

Community Participation

• Try the tool and submit issues for feedback.
• Contribute code to expand search sources and features.
• Share use cases to improve documentation.

Summary

onesearch represents a trend toward specialized, standardized AI toolchains. Its CLI-first unified interface lowers the barrier for AI Agents to access external info and provides an efficient research tool for terminal users. As the AI Agent ecosystem matures, such foundational tools will grow in importance.