Search Candy: Give AI agents passages they can cite
What is Search Candy?
SearchCandy offers Temporal Graph Memory (TGM), a retrieval engine for AI applications. You ingest text or Markdown documents, which are split into sentence-level passages that keep their exact bytes and linked into a knowledge graph by meaning. A query returns the original passage along with the file it came from, and responses include a not_covered flag when no supporting text exists. The engine uses embeddings and does not include a generative LLM, so your application decides how to present or reuse the retrieved context. Every graph carries versioning, so you can query a saved version and see the passages it returned at that time. The documented evaluation reports 152 of 161 answered questions graded excellent, 55 of 55 absent-entity questions refused, no fabricated quotes, and a 22 ms median engine search time. Access is currently a concierge pilot: a 30-minute call to discuss your documents and questions, then account and API key setup during onboarding.
Focus areas: Agent Memory.
Who is Search Candy for?
- Developers building AI applications that need citable document passages
- Teams that need version history on retrieved document content
- Engineering groups evaluating retrieval before committing to a vendor
Is Search Candy worth it?
A focused retrieval API for teams that want sourced, versioned passages rather than generated answers, with access gated behind a pilot conversation
Where Search Candy is strong
- Returns word-for-word passages with a named source file
- Includes a not_covered signal for questions without supporting text
- Versioned graphs let you query and replay earlier document states
Where it falls short
- Documented ingestion accepts only plain text and Markdown; other formats are converted during onboarding
- Queries pause while a graph change indexes
- Pricing is not published and access requires a pilot call
How do you get started with Search Candy?
- 01Book a 30-minute call describing your documents and representative questions
- 02Agree on pilot scope, access, pricing and evaluation criteria
- 03Receive your account and API key during onboarding, then install the SDK and run the quickstart
Search Candy: frequently asked questions
Does SearchCandy generate a final answer?
No. The retrieval engine returns original passages from your documents, and your application decides how to present them or pass them to a language model. Any downstream generated answer needs its own evaluation.
Which file types can I ingest?
The documented ingestion flow accepts plain text and Markdown. Other formats, including PDF, are converted with you during onboarding.
Can I start right after installing the SDK?
No. You need a pilot account and API key first, which are set up during onboarding after a call. Installing the package alone does not grant API access.
What happens to queries while I update documents?
Queries pause while a graph change indexes, and the pause depends on the size of the change. Saved versions remain part of the graph history.
Search Candy alternatives
The closest tools in our directory, matched on category, focus areas and audience.
| Tool | Best for | Free plan | Starting price | Focus areas |
|---|---|---|---|---|
| Search Candy | Developers needing citable, versioned document passages for AI agents | No | — | Agent Memory |
| Versuno | Unify all your AI assets in one place. | Yes | — | Agent Memory |
| Supermemory | Developers adding persistent, cross-model memory to AI agents | Yes | $19/mo | Agent Memory, MCP |
| Vectorize | Teams giving LLM agents memory that improves with use | Yes | — | Agent Memory, Vector Search |
| Cognee | Cognee is an AI platform that builds a dynamic, self-organizing knowledge graph from enterprise data, enabling deep contextual… | Unknown | — | Agent Memory |
| Zep | Chatbot platform with language comprehension and memory. | Unknown | — | Agent Memory |
| Ducky | Developers adding managed search and RAG to AI products | Yes | — | — |
| Julep AI | Developers automating multi-step AI and browser tasks | Unknown | — | — |
| Memable | Personal knowledge organization & retrieval. | Yes | — | — |
| xmem | Xmem centralizes company knowledge, streamlines sharing, and integrates with APIs & LLMs. | Unknown | — | — |
| Mixpeek | Developers searching and classifying large media libraries | No | $25/mo | — |

