uberblickGitHub

Uberblick / Compare

Which tool fits your team?

Uberblick is a narrow tool: the plans, decisions and numbers behind one codebase, for a small or mid-sized team of people and coding agents. Here is what the alternatives do better, what Uberblick does better, and when switching is worth it.

What each tool does better than Uberblick.

These are not consolation prizes. Each is the reason that tool exists, and Uberblick does not compete on it.

Notion

The best general-purpose workspace there is: pages, databases with views, templates for every job, granular sharing, a mobile app, and an AI layer across all of it. Teams can run a whole company in it. Its API and MCP connector let agents read and write pages too.

Confluence

Enterprise permissions, spaces, audit history and the Jira integration. If your issues are in Jira and your organisation has hundreds of people, Confluence is where the documentation belongs.

Google Docs

Real-time co-writing that just works, suggestion mode, comments that people outside your team can answer from an email, and the Workspace suite around it. For prose written with non-engineers, nothing is simpler.

Obsidian

Plain Markdown files you own, instant search over thousands of notes, the graph, canvas, and a plugin ecosystem larger than most products' feature lists. It is the best personal knowledge tool, and sync and publish are available when you want them.

Jupyter

Code and its output in one document, rerunnable, with kernels for Python, R, Julia and more. It is the lingua franca of data work, and tools like nbconvert and JupyterLab make it a complete environment.

Markdown in the repo

No service, no account, no sync problem. Docs change in the same pull request as the code, with the same review. Every agent and every editor can read and write it today.

Where Uberblick is better.

Agents are editors, not integrations

MCP is the front door. An agent reads the sidebar, searches, edits one block, comments on a range and answers your comment with the same tools you use in the browser. No API glue, no export step, no "AI feature" bolted on.

Decisions are records

A decision has an answer, a reason and the issues it affects, and a new record supersedes the old one on the same topic. An agent asks find_decisions for the issue it is working on before it changes anything. In a wiki, the decision is a paragraph somebody may find.

The numbers live in the document

Records are stored in the page; tables and charts read them. When an agent updates the records, the chart updates. No screenshot of a dashboard, no link to a notebook that has since changed.

Local first, always offline

Every computer has a full replica. The editor and the MCP server work without a network, and the hub you run syncs when there is one. Nothing leaves your machines unless you decide it should.

One workspace per codebase

The project directory's .uberblick.json chooses the workspace, for people and agents alike. An agent started in a project cannot write into the wrong corpus, because it never chooses one.

Comments where the text is

Threads sit beside the passage they are about, for people and agents alike. A review happens in the document, not in a chat that scrolls away.

When it's worth it.

Worth the switch when most of these are true:

  • Coding agents work in your codebase, and you want them to read the plan and the decisions before they build, not just the issue.
  • The team is a few people to a few dozen, and the documents are about the product and the engineering behind it.
  • You want the documents on your own machines, and running one Docker container for the hub is acceptable.
  • The repo's Markdown folder has gone stale, or a Notion space has drifted away from what the code does.

Not worth it, today, when:

  • The content is company-wide or non-engineering. Keep Notion or Confluence.
  • You need a hosted service, a mobile app, Windows, or enterprise permissions. Uberblick is at 0.4, on Homebrew for Apple Silicon macOS and Linux x86_64, and you host it yourself.
  • Markdown in the repo still works for you. It is a fine place to be; come back when it stops working.

Side by side.

"Agents" means how a coding agent reads and writes the documents.

 NotionConfluenceGoogle DocsObsidianJupyterRepo MarkdownUberblick
Built forAny team, any contentLarge organisations on AtlassianCo-writing proseOne person's knowledgeReproducible analysisEngineers, in the repoProduct and engineering teams with agents
Where it livesNotion's cloudAtlassian cloud or your data centreGoogle's cloudFiles on your diskFiles on your disk or a hosted kernelGitA replica on each computer, synced by a hub you run
OfflineLimitedNoWith the browser extensionYesYesYesYes, always
Co-editingReal timeReal timeReal time, with suggestionsSingle user; sync as add-onPer file; real time in JupyterLabThrough pull requestsReal time, with comments beside the text
AgentsAPI, MCP connectorAPI, MCP connectorAPIEdit the files directlyRun the cells; edit the filesEdit the files; commitMCP first: search, read, edit one block, comment, update data
Data in documentsDatabases, viewsTables, macrosTables, linked Sheets chartsPluginsAnything computableStatic tablesRecords in the document; tables and charts read them
DecisionsA page or database you designA decision templateA docA noteNot its jobAn ADR fileA record with an answer, a reason and the issues it affects
PermissionsPer page and workspaceEnterprise gradePer documentFilesystemFilesystem or JupyterHubRepository accessWorkspace membership on the hub
Cost modelPer seatPer seatPer seatFree; paid syncFree, open sourceFreeFree, open source; you host

Facts about other tools reflect their public documentation at the time of writing. Tell us where we are wrong and we will fix it.