Show HN: Mantic.sh – Search 480k Files in 0.46s without Embeddings

Why It Matters

By eliminating the need for costly vector searches and cloud APIs, Mantic accelerates AI‑assisted development while preserving privacy and reducing operational expenses for engineering teams.

Key Takeaways

• Sub‑500 ms search across 480k‑file monorepos.
• No embeddings or external services; runs locally.
• Cuts token usage up to 63 % per query.
• Zero licensing cost versus $110k SaaS annually.
• v1.0.12 adds reliable scoring and fixes critical bugs.

Pulse Analysis

Mantic.sh’s core innovation lies in treating codebases as structured metadata rather than raw text. By leveraging Git‑aware file enumeration and a path‑based scoring algorithm, the engine sidesteps the heavy compute and storage demands of traditional embedding pipelines. This architectural shift enables deterministic rankings in sub‑half‑second windows, even for monolithic projects like Chromium, where conventional vector search can take ten seconds or more. The result is a lean, deterministic tool that integrates seamlessly with AI agents without adding latency.

From a business perspective, the speed and token efficiency translate into measurable cost reductions. The benchmark showing a 63 % drop in token consumption means that large language model calls become cheaper, while the sub‑500 ms response time keeps developer workflows fluid. Because Mantic runs entirely on‑premise, organizations avoid data‑exfiltration risks and the recurring SaaS fees that can exceed $100 k annually for a 100‑engineer team. The free AGPL‑3.0 license further lowers barriers to adoption, making it attractive for both startups and enterprises seeking a local‑first AI‑coding assistant.

In the broader market, Mantic positions itself against a wave of embedding‑centric search services that dominate the AI tooling landscape. Its zero‑dependency model appeals to teams wary of vendor lock‑in and API rate limits, while the recent v1.0.12 stability fixes address early‑stage reliability concerns. As AI agents become more autonomous, tools that can quickly surface relevant code without external calls will be critical for scaling developer productivity. Mantic’s roadmap, which includes deeper impact analysis and tighter integration with agents like Claude Desktop, suggests it could become a foundational layer for next‑generation code‑centric AI workflows.