Humata
ProductivitySearch, summarize and compare uploaded documents with source-linked answers. Humata adds team-oriented document access and page-based plans for ongoing knowledge work.
Questions across filesCited passages
PERSONAL KNOWLEDGE / ASK YOUR NOTES BETTER QUESTIONS
Khoj
A self-hostable personal AI assistant for searching documents, asking contextual questions and configuring research workflows, with local-model and cloud-provider options.
Still frame from the official Khoj demonstration showing its web interface; not a FindGoodAI task runExplore the image ↓
What you can do with it
What this tool is for
Khoj is a personal AI application that can combine conversations with document search and online research. Its documentation describes local and cloud model choices, multiple clients and configurable agents. For this listing, the ongoing free route is the open-source self-hosted application.
It can help when useful information is scattered across notes and documents, but it needs deliberate scope. Choose which files enter the system and what kind of question you expect it to answer. A retrieval assistant is most useful when you can inspect its sources and distinguish your material from information found online.
01
Recover an old decision
02
Study a document set
03
Create a focused research assistant
04
Prepare a recurring reading brief
A closer look at the product
Official product image. Click to inspect the details.
Product interface and examples
Key features, practical tasks
Capabilities in practice
- Document-grounded questions: Ask about supported notes and documents that you have added or synchronized. Check whether the retrieved content actually answers the question before accepting the generated explanation.
- Semantic search: The project describes finding relevant documents by meaning. This can complement exact phrase search, particularly when you remember an idea but not the original wording.
- Local and cloud models: The self-hosting guide includes supported local inference and commercial provider connections. Decide which route fits the material and resources; application location alone does not determine where inference happens.
- Several client paths: Web access is available, with documented desktop and note-taking integrations. Configure each client to the intended server so a convenient interface does not silently change the data destination.
- Custom agents: Agents can combine selected knowledge, instructions and model choices. Use a narrow role with clear evidence requirements instead of expecting a persona label to guarantee expertise.
- Repeated research workflows: The documentation describes scheduled queries and email delivery. Define the topic, frequency and recipients deliberately before enabling a recurring external action.
- Optional capabilities: Voice, code execution and image generation appear in the feature catalog. Their dependencies and enabled state should be checked individually rather than assuming a minimal installation includes every capability.
Get started, step by step
- 01
Choose a deployment path
- 02
Configure the inference route
- 03
Add a bounded knowledge set
- 04
Refine the workflow
A practical first workflow
A practical first workflow
- Choose a deployment path. Follow the current Docker or package instructions for your platform. Configure the local instance and credentials before connecting additional clients.
- Configure the inference route. Select a supported local model or an external provider and understand the cost and data path. Confirm the model’s requirements fit the available hardware.
- Add a bounded knowledge set. Import a few permitted documents and try questions with known answers. Check source retrieval before expanding synchronization to more folders or services.
- Refine the workflow. Keep useful instructions and record where errors occur. Add agents or repeated queries only after the basic document search and answer review are working as intended.
Example prompt or task: Find the notes explaining this project decision. Show the relevant sources and their dates, separate the recorded decision from your interpretation, and say when the documents do not establish an answer.
This is an editorial example. It is not a measured retrieval benchmark or a claim that all Khoj deployments use the same model and tools.
Find the right plan for you
Free self-hosted software
Explore official plans ↗Free software Hosted services and model usage are separate
This listing qualifies the local open-source route. It does not assume an unlimited free allowance in the separately operated cloud application.
What is free and what can cost extra
Khoj’s AGPL-licensed code provides a free self-hosted software route, including documented local-model configuration. Hardware, disk space, electricity and maintenance remain the operator’s responsibility.
The cloud application and enterprise offerings are separate from that route. Model APIs, online search, email delivery, transcription or other integrations can introduce additional conditions or charges. Check the exact enabled combination instead of transferring the free software label to every connected service.
A few things worth knowing
Frequently asked questions
Is Khoj free?
The open-source self-hosted application is free. Cloud services and resource costs are separate.
Can it use local models?
The official guide includes local and compatible model-server paths.
Does self-hosting guarantee offline operation?
Only a suitably configured local workflow can avoid remote inference; web research and connected services still need their own network paths.
Can I connect my notes?
Supported document and note-taking integrations are documented. Choose the scope deliberately.
What is a custom agent?
A configured combination of instructions, knowledge and other supported settings for a particular task.
Will scheduling send information externally?
Documented automations can deliver results by email. Review destination and scope before enabling them.
Is Khoj the same as Pipali?
The project links Pipali as another product. This listing describes Khoj, not that separate application.
The details behind your decision
Explore platforms, inputs and outputs, licensing, and access requirements.
Four useful starting projects
Four useful starting projects
- Recover an old decision: Search a permitted project-notes collection for the reasoning behind a past choice. Read the source date and surrounding discussion before treating it as current guidance.
- Study a document set: Ask for explanations tied to selected course or reference material. Save questions that remain unanswered so the assistant does not blur gaps with plausible general knowledge.
- Create a focused research assistant: Configure an agent around a specific topic and a small knowledge set. Ask it to separate document evidence from web findings and unsupported inference.
- Prepare a recurring reading brief: First run a manual query and check its usefulness. Only then consider scheduling it, with explicit delivery settings and a plan for outdated or irrelevant results.
Platforms, inputs, outputs and setup
Platforms, inputs, outputs and setup
Self-hosting instructions cover container and Python package options, with platform-specific requirements. A browser can connect to the server, while additional clients require their host settings to point to the appropriate instance.
Inputs can include supported documents, synchronized notes and questions. Outputs include retrieved passages and generated responses, with optional capabilities depending on configuration. Keep a record of data sources and model settings when using answers for ongoing work.
Data, permissions and licensing
Data, permissions and licensing
Review what synchronization includes and where model requests go. A local server can still use a cloud model, and an online search query may disclose details from the research question.
Configure authentication before remote access and avoid sharing conversations containing private source material. Treat generated answers as drafts that should be checked against the underlying notes before becoming decisions or published claims.
Sources, documentation and help
Official sources and review date
Official sources and review date
Reviewed October 3, 2026. Product facts come from the official sources below. Suggested projects, prompts and review methods are FindGoodAI editorial guidance, not measured performance results.
Your experience can help someone choose
COMMUNITY
Sign in to rate and comment
Tried it? Share your experience
Loading experiences…
Manage my comments
Community and moderation
Saved tools are private. Approved comments are public; edits return to moderation. Comment counts include approved comments and replies. Each person has one rating. New ratings are automatically approved; withdrawn, unapproved or invalid ratings do not count. Share real experiences and avoid spam, private information or personal attacks.
For moderation appeals or data requests: support@findgoodai.com
NEXT STEP
Explore Khoj ↗Make a small collection of notes easier to use
Bring a real task and see how it fits the way you work.
Jump to a section
User experiences