TinySearch
Search, crawl, chunk, and rerank the web into compact source-grounded evidence for MCP agents.
- Source-attributed snippets
- Reranked against the question
- No API key required
Open-source local agent infrastructure
Find the evidence.
Remember what matters.
TinySuite gives local agents two disciplined context loops: TinySearch finds compact, source-grounded web evidence; TinyContext recalls only the durable memory that fits.
- 230stars
- MITOpen source
- LocallyRuns
- MCPProtocol
01 — The tools
Give agents better context without making them heavier.
TinyContext
Store concise local memories, retrieve them with hybrid ranking, and return only what fits the token budget.
- Hybrid keyword + vector recall
- Hard token budget per call
- Everything stays on your machine
02 — Install
Set up the full suite from your terminal.
Pick your coding agent. One command registers both MCP servers.
bash
claude mcp add --scope user tinysearch -- uvx --from "tinysuite-search[server]" tinysearch && claude mcp add --scope user tinycontext -- uvx --python 3.12 --from "tinysuite-context[server]" tinycontextRequires uv and the selected agent CLI. The first run downloads each tool’s local dependencies.
03 — Why it matters
Context is a budget. Most agent stacks treat it like a landfill.
Whole pages, old transcripts, and every stored fact crowd out the information an agent actually needs. TinySuite retrieves selectively and keeps each step inspectable.
- Context dump
- Raw pages and history enter the prompt
- Relevance is hard to inspect
- Local models run out of room
+ TinySuite
- Ranked evidence and memory only
- Local-first, no account wall
- Every selected item has a reason
04 — Shipping
What shipped lately.
Recent releases across TinySuite. Open a release for its full changelog.
TinyContext v0.4.1
What changed Add the list memories MCP tool for newest first, filterable, paginated memory previews. Add the get memory MCP tool to retrieve a memory by full ID or short reference. Keep the FastAPI endpoints aligned w…
TinySearch v0.6.2
TinySearch v0.6.2 TinySearch v0.6.2 makes web research more focused and efficient, with smarter handling of overlapping crawl evidence and a set of quality and performance optimizations. What's changed Semantic dedupl…
TinySearch v0.6.1
Full Changelog : https://github.com/TinySuiteHQ/TinySearch/compare/v0.6.0...v0.6.1
05 — Writing
Field notes on building token-light agents.
Practical guides and design tradeoffs for local AI workflows.
What We Learned From Benchmarking Web Research Agents
We built an agent-in-the-loop benchmark to compare web research systems. The scores were useful, but the traces taught us much more about retrieval, context, latency, and tool design.
Link Discovery Without a Crawler: TinySearch 0.6.0
Scraping needs URLs you already have. Crawling follows links on its own budget. TinySearch 0.6.0 adds bounded related links so the agent picks the next page, plus support for an external browser over CDP.
TinySearch vs. Tavily, Exa, Perplexity Sonar, and Brave Search API
How TinySearch's self-hosted, no-key research pipeline compares to Tavily, Exa, Perplexity's Sonar API, and Brave's Search API for AI agents.
Get started
Two tools. One discipline. No account wall.
Install both MCP servers in a single command and give your local agent context worth reading.