Warning
Alpha software. Needle-X is still in active development and test. Install flow, local state layout, CLI details, and output shape may still change.
Turn messy web pages into compact, proof-carrying context for AI agents.
Smaller packets. Fewer hops. Real provenance.
- Smaller output Needle-X returns much less context than extraction-heavy tools.
- Source-backed It carries proof, not just extracted text.
- Less cleanup A downstream agent does less work before it can act.
| Metric | Needle-X | Tavily | Jina | Firecrawl |
|---|---|---|---|---|
| Avg packet bytes | 4436 | 6975 | 30565 | 72166 |
| Claim-to-source steps | 1 | 2 | 2 | 2 |
| Post-processing burden | 0.25 | 1.92 | 1.86 | 2.50 |
| Proof usability | 1.0 | 0 | 0 | 0 |
Needle-X vs Jina:
- about 85.5% smaller packets
This is the current sweet spot:
- compact context
- direct verification
- low-friction agent consumption
Needle-X includes local Discovery Memory backed by SQLite.
The story is simple:
- first run observes and compiles
- later runs reuse local verified evidence
- repeated use improves local retrieval without hosted infra
Current warm-state result:
- 30/30 selected-url correctness
- 30/30
discovery_memoryprovider selection
Guardrail:
- warm-state local retrieval claim
- not a cold-state open-web seedless claim
readquerycrawlproofreplaydiffmemory stats/search/prune
Default output is AI-first:
- compact packet first
- proof inline when useful
- full diagnostics only on demand
- browser-like fetch by default for real-world targets
- MCP server accepts both standard
Content-Lengthframing and raw newline-delimited JSON
needlex read https://example.com --json
needlex query https://example.com --goal "pricing" --json
needlex proof proof_1 --jsonLinux and macOS:
curl -fsSL https://raw.githubusercontent.com/Josepavese/needlex/main/install/install.sh | bashWindows:
irm https://raw.githubusercontent.com/Josepavese/needlex/main/install/install.ps1 | iexInstalled command:
needlex
This installer downloads the right release binary. Full details:
- browser agent
- search engine
- generic scraper
- LLM-first reader


