Workdocs Canvas

Workdocs are persistent Markdown documents built into your workspace, serving as a shared canvas between you and the agent for software design, architectural specifications, and implementation tracking.

The Role of Workdocs

In standard conversational AI interfaces, requirements and architectural decisions frequently scroll away as sessions grow longer. Workdocs solve this by introducing an anchor document that persists across the lifespan of a feature:

• Bidirectional Collaboration: Both you and the agent can write, edit, and organize sections of the canvas simultaneously.
• Continuous Context Grounding: An open Workdoc is automatically injected into the agent's context, ensuring that all subsequent code suggestions follow the agreed-upon design.
• Versioned Project Artifacts: Workdocs can be saved directly into your repository (e.g. docs/rfcs/auth-spec.md) so that team members can review them in pull requests.

Working with Workdocs

1. Generating Specifications

When embarking on a complex task, start by having the agent outline the architecture before writing code:

Draft a technical design in our Workdoc for migrating our session store from in-memory maps to Redis. Include failure modes and data migration steps.

The agent will populate the canvas with system diagrams, interface definitions, and a checklist of files to modify.

2. Refining with Inline Comments

You can highlight specific paragraphs or code blocks within the Workdoc to leave inline comments:

• Clarify constraints: "We cannot use Redis Cluster here; use single-instance Redis Sentinel."
• Request alternatives: "Provide an alternative fallback path if Redis is unreachable."

The agent reads these targeted annotations and updates the corresponding sections of the document.

3. Tracking Execution Checklists

Workdocs support standard Markdown checkboxes (- [ ]). As the agent implements each module, runs tests, and verifies changes, it checks off completed items in the Workdoc, providing a clear visual record of progress.