Design, canvas & AI
Get answers in AI chat grounded in your team's docs and code context, and choose a model and reasoning effort.
Specify chat is a workspace for answering questions and creating documents or canvases from workspace knowledge, connected code, and external sources. Describe both the subject and the result you want.
Ask with specific sources
Use @ in the input to reference a document or repository. For example, ask the agent to compare the login implementation with the authentication guide. The connected GitHub App must have repository access.
Check the order processing flow in this repository.
List differences from the current documentation with code evidence,
and keep anything you could not verify in a separate list.

Turn answers into work
After a summary or explanation, request a new document, edits to an existing document, or a canvas. Specify the save location and scope. For existing documents, identify sections to preserve.
The task list can show progress for reading code and reflecting it in a document separately. After the document update completes, open the output to verify the saved content.
Discuss the design while drawing
Ask “Draw this flow as a diagram” to visualize the structure. Open a canvas and continue the conversation in Ask to edit this canvas… at the bottom to refine layout, labels, and connections. Selected shapes or connections are also supplied as context.
Follow Architecture diagrams and canvases for whole-design commands and focused edits.
Check evidence
Open the conversation's evidence to inspect the sources used. You can retract incorrect or unsuitable evidence; retracted evidence is excluded from later answers. Evidence may also change state when its source changes or becomes inaccessible.
Before an important decision, compare the answer with the original material and code behavior. If the agent could not read the full scope, identify the remaining areas and request further checks.
Model and reasoning effort
Use the input's model settings to choose a model and its supported reasoning effort for the current conversation. Model changes are locked during a response. Higher effort can increase response time and usage. See the measured evidence in the model selection guide.
External services and coding agents
Connect services and enable AI access to include permitted external information. To use the same team documents in your IDE, connect Specify MCP.