Notion alternative · for developers using AI agents

The Notion alternative for developers who build with AI agents.

Notion holds specs for people to read. ContextsBase holds features, business rules, data model, and test cases in a form Claude Code, Cursor, and Copilot read directly over MCP. Nothing pasted into chat.

  • Works with any MCP client
  • Bring your own agent
  • Free for one project
notion · Password reset (page)

Password reset

Users should be able to reset their password from the sign-in screen. We email them a link. It should probably expire after a while (15 min?) and only work once. Check with Sam about the enterprise tier.

  • Build the form
  • Send the email
Select all · copy · paste into every new chat

you › “Here’s the spec again. Ignore the Sam part…”

contextsbase · F-3 Password reset
F-3Password reset
Published
Functionality
User requests a reset link from the sign-in screen and sets a new password.
Business rules
  1. R1Reset link expires 15 minutes after it is issued
  2. R2A link can be used exactly once
Entities
UserResetToken
Test cases
T-7T-8
read by your agentnext_iteration_item

Left: what you paste into chat today. Right: what the agent pulls on its own.

Why Notion stops working the day an agent joins the team.

// notion page says

“…customers on the enterprise tier can…”

// agent wrote

if (customer.tier === 'enterprise')

Customer has plan_id, not tier

A page is prose, not a schema

The agent reads a sentence and guesses a field name.

Notion pagecopychat

you › “Here’s the spec…”

page edited yesterday · the paste is from last week

The agent gets a copy, not the source

You paste. The page moves on. The paste does not.

Password reset · ticked by someone, sometime

F-3Implemented·T-7T-8passed

recorded by the agent, against the spec it read

Nothing records what got built

A checkbox says done. It cannot say which tests passed.

Notion vs ContextsBase for developer teams using AI.

Notion is a workspace for people. ContextsBase is the knowledge base for the agents that write your code. Here is where they differ.

Feature by feature comparison of Notion and ContextsBase for developers who work with AI coding agents
AspectNotionContextsBase
Built forPeople reading and writing pages.Coding agents reading specs, and the people who write them.
Feature specsFreeform pages and databases.Structured features (F-n) with functionality, business rules, and edge cases. Draft until published.
Data modelTables and relations you lay out by hand, for humans to browse.Entities with fields and foreign keys, served to the agent with the spec.
Test casesChecklists and to-dos.Given, When, Then cases (T-n) that assert business rules. Agents turn them into Playwright tests and record the result.
Agent accessNotion API and Notion MCP return page content as blocks of text.Native MCP server. One call returns spec, entities, flow, tests, and theme tokens.
Work queueKanban boards people move cards on.Iterations (I-n) with atomic claims. Two agents never take the same step.
Implementation recordA checkbox someone ticks.The agent records what it built and which tests passed. Editing the spec later marks the record Outdated.
Design tokensNot applicable.Theme Designer publishes colors, type, radius, and spacing for agents to read.
AINotion AI drafts and summarises for the reader.No agent included. Bring Claude Code, Cursor, Copilot, or Windsurf.
PricingPer seat.Free for one project. Premium at $5 per month.

Notion is a trademark of Notion Labs, Inc. ContextsBase is not affiliated with Notion.

What replaces the Notion page.

Every piece a coding agent needs, as a first-class object with an ID it can ask for.

  • F-n

    Features

    Functionality, business rules, edge cases. Draft until you publish.

  • Entities

    Data model

    Fields and foreign keys. The agent stops guessing column names.

  • Flows

    Decision trees

    Branches of a user journey, so the agent builds every path.

  • T-n

    Test cases

    Given, When, Then. Turned into Playwright tests that assert the rule.

  • I-n

    Iterations

    Ordered features. Agents pull the next one, one at a time.

  • Theme

    Theme Designer

    Colors, type, radius, spacing. Published tokens the agent applies.

  • Edits

    Page Editor

    Click an element on your live page, change it, save. The agent ships it.

  • MCP

    Five tools

    list_iterations, next_iteration_item, get_feature_spec, and two record calls.

Move one spec out of Notion. Ten minutes.

You do not migrate a workspace. You move one feature and watch what the agent does with it.

  1. 1

    Paste the spec

    Create a feature. Split the Notion page into functionality, business rules, and edge cases. Add test cases as Given, When, Then. Publish.

    F-3 · Published
  2. 2

    Queue it

    Add the feature to an iteration and open it to agents. Order decides what gets built first.

    I-1 · step 1
  3. 3

    Point your agent at it

    One config block in Claude Code, Cursor, or Copilot. The agent calls next_iteration_item and starts.

    next_iteration_item
~/.claude/mcp.json · or your agent's equivalentone block
{
  "mcpServers": {
    "contextsbase": {
      "url": "https://app.contextsbase.com/api/v1/mcp",
      "headers": { "Authorization": "Bearer mcp_live_…" }
    }
  }
}

Tokens are scoped to read, write, or record. The free plan includes one per project.

Keep Notion for people. Give agents a base of their own.

Most teams run both. The difference is what the agent reads.

Stays in Notion

Moves to ContextsBase

  • Meeting notes
  • Roadmap and OKRs
  • Company wiki
  • Customer research
  • Hiring and onboarding docs
  • Feature specs
  • Business rules
  • Data model
  • Test cases
  • Work queue for agents

Free for one project.

Free

Free forever

$0

One real project, fed to your agent.

  • 1 project
  • 100 features
  • 3 members
Create free project

Premium

Coming soon

$5per month

Cancel any time.

  • Unlimited projects
  • Unlimited features
  • Unlimited members
Coming soon

Questions developers ask before leaving Notion.

Is ContextsBase a full replacement for Notion?

No. It replaces the part of Notion that developers use to hold specs for AI coding agents: feature specs, business rules, the data model, test cases, and the work queue. Meeting notes, roadmaps, and the team wiki can stay in Notion.

Which AI coding agents work with it?

Any Model Context Protocol client: Claude Code, Cursor, GitHub Copilot, Windsurf, and others. ContextsBase does not ship its own agent. You bring the one you already use and point it at your project with one config block.

How is this different from Notion AI?

Notion AI writes and summarises for the person reading the page. ContextsBase serves the spec to the coding agent that writes the code, then records what the agent built and which tests passed.

Can I move my existing Notion specs over?

There is no one-click importer yet. Paste a spec into a new feature and split it into functionality, business rules, and edge cases, then add test cases as Given, When, Then. Most teams start with a single feature and one iteration.

What does the agent actually receive?

One call to next_iteration_item returns the published spec, its entities and fields, the flow, the test cases, and the current theme tokens. Nothing is copied between tabs or pasted into chat.

What does it cost?

Free for one project with 100 features, 3 members, 1 MCP token, and unlimited open iterations. Premium at $5 per month adds unlimited projects, features, and members, 10 MCP tokens per project, and 20 open iterations.

Your agent has never read your Notion. It will read this.

One feature, one iteration, one line of config. Free for one project, no agent required.