Khoj

Agents are documented as configurable with tools including web search and a code execution sandbox. Document ingestion covers local files and folders synced through desktop, Obsidian and Emacs clients, a Notion workspace synced from the web app, and direct upload, across PDF, Markdown, org-mode, Word, Notion and image formats, with a dedicated Data Sources documentation section. No catalog of third party application connectors for CRM, ticketing, messaging, warehouse or calendar systems is documented, and neither is Model Context Protocol tool support.

Sourcedocs.khoj.dev setup sync-your-knowledge, and clients documentation, github.com/khoj-ai/khoj READMEread 2026-08-31

Users can create agents with custom knowledge, persona, chat model and tools to take on any role, schedule automations for repetitive research, and run an autonomous deep research mode that plans, browses and synthesizes results. Intermediate steps including intent detection and web search are model-driven and separately configurable through ServerChatSettings Default and Advanced model fields. A code execution sandbox runs Python as part of a task. No multi-agent coordination, agent-to-agent delegation, conditional branching or subworkflow abstraction is documented.

Sourcegithub.com/khoj-ai/khoj README, docs.khoj.dev setup and navigationread 2026-08-31

Khoj indexes the customer's own documents including PDF, Markdown, org-mode, Word, Notion files and images, with semantic search over them and verifiable citations returned with answers. The desktop, Obsidian and Emacs clients keep local files and folders synced to the server, a Notion workspace syncs directly from the web app, and files can be uploaded directly in the web app. The index is held in the customer's own Postgres database on their self-hosted instance, and can be queried using local models via Ollama, vLLM or LMStudio so no data leaves the machine.

Sourcedocs.khoj.dev setup sync-your-knowledge and documentation, github.com/khoj-ai/khoj READMEread 2026-08-31

Agent code executes inside a sandboxed environment, answers return verifiable citations allowing a user to check sources before acting, and the assistant's search scope can be constrained to reduce fabrication. An administrative panel is gated behind credentials created at first run, and anonymous mode which bypasses login is documented as a deliberate opt-in for local single-user installs. No approval gate, checkpoint, pause for human review, per tool authorization or pending action queue is documented.

Sourcedocs.khoj.dev setup and privacy documentation, github.com/khoj-ai/khoj READMEread 2026-08-31

An admin panel is documented at the server admin path with an administrator account and password created during first run, alongside a separate authentication setup path and an explicit anonymous mode for local single-user installs that bypasses login.

Deployment configuration covers allowed domains, HTTPS behavior and reverse-proxy trusted origins, and the documentation states Khoj is only accessible on the machine it runs on by default with remote access requiring deliberate configuration. A dedicated privacy page documents the data position, and the customer holds the database.

No security certification, attestation, single sign-on or SCIM provisioning is documented, which is expected for a self-hosted open-source deployment where the operator owns the perimeter.

Sourcedocs.khoj.dev setup, privacy and advanced authentication and remote documentationread 2026-08-31

Answers return verifiable citations to the source documents they were grounded in. An administrative panel at the server admin path provides configuration review over the customer's own database, and server log output is described in the setup documentation with the customer holding the Postgres database directly. No agent execution trace, tool call record, run history view, audit log of user actions or analytics capability is documented.

Sourcedocs.khoj.dev setup and privacy documentation, github.com/khoj-ai/khoj READMEread 2026-08-31

Agents are created with their own knowledge, and the document index plus conversation history persist in the customer's own Postgres database on their self-hosted instance, with desktop, Obsidian and Emacs clients keeping files and folders continuously synced. State therefore survives sessions, restarts and upgrades and is owned by the customer.

No documented capability describes the agent updating its own knowledge or instructions from experience, carrying conclusions between conversations, or accumulating learned context over time. A pull request title in the project changelog references disabling memories for users, but the documentation describes no memories feature.

Sourcegithub.com/khoj-ai/khoj README and changelog, docs.khoj.dev setup sync-your-knowledge documentationread 2026-08-31

Khoj is AGPL-3.0 open source and self-hosted, with documented installation via Docker, Docker Compose or pip on macOS, Windows through WSL2, and Linux, backed by the customer's own Postgres database or an embedded database option.

Configuration covers domain, HTTPS, allowed hosts and reverse-proxy settings through environment variables, with separate remote access documentation and auto-start guidance via cron or Task Scheduler. The vendor states data never has to leave the customer's private network and that Khoj can be used without an internet connection when deployed on a personal computer. Khoj Cloud was sunset on 15 April 2026, leaving self-hosting as the only deployment route.

Sourcedocs.khoj.dev setup, privacy and advanced remote documentation, app.khoj.dev sunset noticeread 2026-08-31

Each agent is one the user creates, with custom knowledge, persona, chat model and tools, and the vendor links a blog post with a step by step guide to building custom agents. No template library, preset or prebuilt agent set, shareable agent format, community gallery or starter pack is published. Khoj Cloud, which hosted any shared agent surface, was sunset on 15 April 2026, so a new user begins from an empty self-hosted installation.

Sourcegithub.com/khoj-ai/khoj README, docs.khoj.dev navigation, app.khoj.dev sunset noticeread 2026-08-31

The vendor documents Automations that schedule repeated jobs for Khoj to run a query automatically at a set time and frequency in the user's local time zone and email the research to the user, configurable, shareable and deletable from an automations page, with self-hosted installs needing Resend and authentication for the email. Clients are documented for web, desktop, Obsidian, Emacs and WhatsApp. No inbound webhook or external event trigger is documented.

Sourcedocs.khoj.dev features/automations and clientsread 2026-09-29

Chat models are configured per provider through an AI Model Api object holding the credential, with OpenAI, Anthropic and Gemini documented alongside offline and local models via any OpenAI-compatible server including Ollama, vLLM, LMStudio and llama-cpp-server by setting the API base URL.

Individual chat models are registered with optional tokenizer and max-prompt-size fields and a per-model vision-enabled flag, and users select their preferred model in their own settings. ServerChatSettings holds separate Default and Advanced model choices used for intermediate steps such as intent detection and web search. Agents can be created with their own chat model.

Sourcedocs.khoj.dev setup add-chat-models and advanced ollama and lmstudio documentation, github.com/khoj-ai/khoj READMEread 2026-08-31

The server, clients and docs are open source under AGPL-3.0 in one public repository, and the desktop, Obsidian, Emacs and WhatsApp clients talk to a self-hosted server over HTTP at a configurable URL, so the product can be read, modified and self-run. The documentation has no API reference, SDK or MCP server page, so no documented interface exists for a customer's own code to invoke Khoj agents or automations. Open source lets a customer change the product rather than call it.

Sourcedocs.khoj.dev sitemap and github.com/khoj-ai/khojread 2026-09-29

No testing, evaluation, scoring, benchmarking or regression capability is described in the documentation or the project README. No self-checking or verification gate on agent output is documented either. Verifiable citations let a human reader check an answer's sources after the fact.

Sourcegithub.com/khoj-ai/khoj README, docs.khoj.dev navigation and setup documentationread 2026-08-31

Khoj provides a code execution sandbox in which agents run Python, and web access through retrieval and semantic search over fetched content. No browser control, navigation, form filling or screen interaction is documented. Clients are documented for web, desktop, Obsidian, Emacs and WhatsApp, all of which are interfaces to Khoj rather than interfaces Khoj operates.

Sourcedocs.khoj.dev setup and clients documentation, github.com/khoj-ai/khoj READMEread 2026-08-31