Dexiask offers a robust suite of features designed to address the common pain points of modern software development teams. The ability to index a repository once and then ask questions in plain language is a significant time-saver, eliminating the need for constant grepping and navigation through call sites. The semantic and lexical code search, powered by Qdrant vectors and Voyage embeddings, ensures that results are both relevant and comprehensive. The optional LLM-generated domain docs provide additional context, embedding architecture and module descriptions alongside the source code. This creates a unified view where one query can return both documentation and source references. The memory feature is particularly impressive, as it consolidates observations from chats into durable memory, making the agent sharper about your specific codebase over time. Multi-user support with GitHub sign-in and enforced per-user repo access adds a layer of security and collaboration suitable for enterprise environments. The extensibility through SKILL.md packs and MCP servers allows teams to customize the agent to their specific conventions. Overall, Dexiask presents a compelling, self-hostable solution that keeps your code and keys within your own infrastructure.
Questions about Dexiask
What is Dexiask?
Dexiask is an open-source, self-hosted AI tool for querying a codebase that indexes a repository once and allows users to ask questions in plain language. It uses semantic and lexical code search powered by Qdrant vectors and Voyage embeddings, and includes a read-only Claude agent that can read files and reference the code index to produce answers with file and commit citations. The product is MIT licensed and runs locally via Docker, combining an engine, indexer, memory store, Qdrant, Postgres, and a web interface.
Who is Dexiask intended for and what problem does it solve?
The tool is designed for teams that want to eliminate stale documentation and tribal knowledge bottlenecks by letting anyone ask questions like 'where is auth handled?' or 'how does a request flow?' directly against the code. It solves the problem where knowledge rots everywhere except the code itself, allowing users to bypass the need to grep or wait for a single person to explain a module. The code serves as the source of truth, ensuring answers are grounded in committed code rather than outdated wiki pages.
How can a new user quickly set up and start using Dexiask?
To get started, users can run the quickstart command which clones the repository and sets up the entire stack via Docker Compose, bringing up the web, agent engine, indexer, memory, Qdrant, and Postgres components. Once the web is ready on localhost:3000, users can index a repository by pointing it at any git repo, local or private remote, and then begin asking questions in plain language. The process involves mounting the repo, indexing embeddings and domain docs once per branch, and then answering queries with citations.
How does the Dexiask engine process a user's query from start to finish?
The system operates by mounting a repository, indexing it with embeddings and domain docs, and then answering questions grounded in committed code. When a user asks a question, the read-only Claude agent reads files, searches the web, and calls the code index to stream rich answers with syntax highlighting, diagrams, and math. Users can also interact with follow-up choices, file attachments, and git blame/log/diff features directly from the interface.
What specific capabilities does the read-only Claude agent provide?
Dexiask offers a read-only Claude agent equipped with tools to read files, search the web, and call the code index to produce answers. The agent streams rich content including syntax highlighting, diagrams, and math with clickable follow-up choices, file attachments, and git blame/log/diff at hand. This integration allows users to get detailed, context-aware responses directly within the platform.
How does the semantic and lexical code search feature work?
The search functionality combines semantic and lexical code search using Qdrant vectors and Voyage embeddings over any repo's default branch. The indexing process is incremental and idempotent, ensuring that embeddings and domain docs are updated efficiently without redundant processing. Users can also optionally generate LLM-generated domain docs for architecture, modules, and concepts that are embedded beside the code.
How does the memory feature improve the agent's performance over time?
The memory feature allows the agent to remember observations from chats by consolidating them on a schedule into durable user, repo, and global memory. These consolidated memories are injected into later answers, making the agent sharper about the user and their codebase over time. The memory is also browsable and editable, giving users control over how their interactions are stored.
How does Dexiask handle multi-user access and permissions?
Dexiask supports multi-user collaboration by allowing users to sign in with GitHub and enforce admin/member roles. It manages invites, per-user repo access enforced in the indexer, and per-user chat history. Private repos are handled via one central git token, ensuring secure access management within the team environment.
How can users customize or extend Dexiask to fit their specific needs?
Users can extend the tool by dropping in SKILL.md packs to teach the agent their specific conventions or bringing their own MCP servers for extra tools. The platform also supports asking from the browser or straight from Slack, offering flexibility in how users interact with the system based on their workflow preferences.
Which team structures or development environments benefit most from Dexiask?
The tool is particularly suited for environments where documentation often becomes stale quickly after a service ships, or when tribal knowledge bottlenecks occur when a key person is on leave or on-call. It is ideal for teams that want to avoid spending an afternoon grepping and jumping through call sites to answer questions that the codebase could resolve in seconds.
What are the core steps involved in the Dexiask workflow?
The process involves three main steps: mounting the repository, indexing the codebase with embeddings and domain docs, and then asking questions. The indexing happens once per branch and is incremental after that, ensuring efficient updates while maintaining accuracy in the code index.
What specific features do reviewers praise in Dexiask?
Reviewers have praised the ability to ask plain language questions like 'how do refunds get reconciled?' and receive answers with file and commit citations. The tool effectively addresses the issue of knowledge rotting outside the code, providing a reliable way to access information directly from the source.
What is the official website of Dexiask?
The official website of Dexiask is dexiask.com.
How is Dexiask rated on Trustburn?
On Trustburn, Dexiask has a rating of 4 out of 5 based on 4 reviews.
Post your question publicly — other visitors, and whoever manages this page, can answer. It will be visible on this page.
Want to hear about this company again? Follow it and we’ll email you when a new review appears.
Based on reviews, listing data and the company's website. Verify anything important with the company directly.
It was super easy to get started with the docker compose setup. I had it running in no time and already asking questions about my codebase. Very cool tool!
The Docker Compose setup gets you running in minutes, which is exactly what I needed for my local dev environment.
The concept behind Dexiask is undeniably clever, leveraging Claude to provide instant answers to codebase queries without relying on stale wikis or tribal knowledge. However, from an analytical perspective, the pricing structure remains frustratingly opaque. The website highlights that the tool is MIT-licensed and self-hostable, which suggests a potentially low barrier to entry for individual developers or small teams willing to manage their own infrastructure. Yet, for an organization considering scaling this across multiple repositories or departments, the absence of clear licensing tiers or usage-based pricing models is a significant gap. The mention of requiring Anthropic and Voyage API keys implies that costs are tied to external service usage, but the website does not specify whether Dexiask itself charges a subscription fee on top of those variable costs. This ambiguity makes it difficult to forecast operational expenses accurately. For a tool that promises to reduce the time engineers spend grepping through code, the lack of transparent, upfront pricing information feels like an oversight. Potential enterprise clients need to know if there are additional platform fees or if the cost is purely pass-through based on token usage. Until the financial model is explicitly detailed, this remains a strong technical solution with a significant commercial uncertainty factor that could complicate budget approvals.
Questions & Answers
No questions yet — be the first to ask.
Sign in to post
Your text is saved — you'll come right back to post.
Don't have an account? Create one
News & Press
No news yet. This is where the business shares its own updates and press.
Photos
No photos yet. Photos shared by reviewers will show up here.
About Dexiask
Dexiask is an open-source, self-hosted AI tool for querying a codebase. It indexes a repository once and then allows users to ask questions in plain language, using semantic and lexical code search powered by Qdrant vectors and Voyage embeddings. The product includes a read-only Claude agent that can read files, search the web, and reference the code index to produce answers with file and commit citations. It is MIT licensed and runs locally via Docker, combining an engine, indexer, memory store, Qdrant, Postgres, and a web interface.
- Website
- dexiask.com
- Categories
- Computer software
