Skip to content

Chat and knowledge

Ask about your code. Every claim comes with its source.

Chat answers questions about your team's work from your own sources: your repositories, your knowledge, connected apps, memories, data sources and skills. It does the reading itself and shows exactly where each answer came from.

It opens the code, like a teammate would

Chat looks a repository over (its README, its folders, manifests such as package.json and agent notes such as AGENTS.md), searches the code, reads files by line range and checks the code graph, over as many steps as the question needs. Every read is pinned to one commit. It never asks you to open or paste a file it can read; an answer that does is sent back before you see it.

CoderHelm Chat answering how a web app is put together: a line saying it read two files and searched knowledge, the steps it took, and an answer with numbered citations to source files.
One quiet line says what Chat read; open it to see each step. The answer cites files by number. Example data.

Asking

  • It picks what to read

    Every question can use everything you can see. Name a repository in the question and Chat reads that one.

  • @ to narrow

    Type @ to point a question at one repository or knowledge collection. Nothing is narrowed unless you choose it.

  • Numbered citations

    Hover or tap a number to see the source: the file path and lines with a link to them on GitHub at that commit, or the knowledge page and its date.

  • No guessing

    When nothing in your sources supports an answer, Chat says so instead of making one up.

  • Private until shared

    Your chats are yours. Share one and teammates can read it; only you ask in it.

  • Metered and capped

    Chat's model calls show under their own name on the usage page and count toward your token limits.

Knowledge that stays in sync

A knowledge base holds pages from the apps your team writes in, next to files you upload. Pick what to bring in once and CoderHelm keeps it current, removes what is deleted and cites each page with a link back to where it lives. Agents can search a base too, so a plan can follow your runbook instead of guessing.

SourceWhat it brings in
NotionPages and databases
ConfluenceSpaces or page trees
Google DriveFolders or files: Docs, Sheets, Slides, PDF, Word, Excel and more
SanityPublished documents. Sign in with Sanity, or use an API token for instant updates
GitHubMarkdown and text files on the default branch
WebsitesPublic pages or a sitemap
FilesPDF, Word, Markdown, text, CSV, Excel and HTML uploads
A CoderHelm knowledge base fed by Notion, Google Drive, Confluence, Sanity and uploaded files, each with its sync state.
A knowledge base fed by several apps, each card showing whether it is up to date. Example data.

CoderHelm only ever reads from these apps: nothing is changed, created or deleted there.

  • Tickets to PRs

    Assign a GitHub issue or Jira ticket to CoderHelm and it plans the change, writes it on its own branch, runs your build and tests, reviews its own work and opens a pull request.

  • Skills

    CoderHelm works your way: it reads your AGENTS.md, the skills and agent briefs your team keeps in a repository, the MCP servers you connect and what it learned on past runs.

  • Security

    How CoderHelm handles your code and data: changes on its own branch, builds in a sandbox with no credentials, read only data access, KMS encrypted secrets, signed webhooks and team roles.

Put CoderHelm on your next ticket.

CoderHelm is in open beta. Sign up and bring your own Anthropic API key.