Do I have to change how I work?
Usually, no. Graphode meets the work where it already happens. You can start from Markdown, documents, diagrams, images, spreadsheets, Git repositories, AI chats, agent output or manual editor input, depending on which product surfaces are available.
The point is not to force a new ritual. The point is to turn existing work into a reviewed, source-linked GraphoTree and then use that model as stable memory for future AI-assisted execution.
Can Graphode create a GraphoTree from documents or Markdown?
Yes. Documents, Markdown notes, diagrams, images and other files can be treated as source material for candidate model structure where supported.
Graphode can help create and refine the GraphoTree from those sources with source trace and review. It does not claim that every document is perfectly understood without human correction.
Can Graphode create a GraphoTree from Git?
For engineering work, Git is one of the most important source surfaces. Graphode can use repository bindings, files, branches, commits, changed files and related evidence as input for the model where supported.
The repository remains the repository. Graphode adds a source-linked model around it so services, features, dependencies, rules, tasks, artifacts and acceptance boundaries can be understood as part of one system.
Can I keep designing through ChatGPT, Claude or Codex?
Yes. AI chat can be a primary workflow rather than an import afterthought. A user can keep designing systems, features and implementation plans in a familiar assistant while Graphode stores the durable project reality around it.
Through the app and MCP-style tooling, a compatible assistant can request scoped context, ask for details, propose GraphoTree updates, report progress, register artifacts and create checkpoints.
Can agents populate or update the GraphoTree?
Where supported, yes. An agent can help extract candidate nodes, relationships, constraints, tasks, acceptance criteria and source references from existing material or from an active work session.
Important changes remain reviewable. The public promise is candidate creation, reviewed proposals and source-linked refinement, not blind autopublish.
Is Graphode a memory layer across providers and models?
Yes, that is one of the strongest public explanations. Graphode is not memory as a long chat transcript. It is structured memory: entities, relationships, decisions, constraints, source trace, checkpoints, artifacts and acceptance state.
That structure can survive provider changes, model switches, new sessions and handoffs between agents and people.
What happens if I switch models or providers?
Graphode is provider- and model-agnostic by design. If a user needs to move from one assistant or model to another, Graphode can provide the modeled reality, active task scope, latest checkpoint, open gaps and acceptance criteria to the next compatible tool.
That means the user can continue from the last meaningful state instead of rebuilding the project story from scratch.
What are checkpoints?
A checkpoint is a meaningful state in a long run: what has been done, what evidence exists, which acceptance criteria may be covered, what is still open and what the next safe action should be.
Checkpoints make agent work resumable and auditable. They also give Graphode something to compare against the model when checking whether execution has drifted.
How does Graphode control drift?
Graphode can compare progress, checkpoints, artifacts and acceptance state against the modeled reality. If the work no longer matches the task boundaries, source evidence or constraints, that drift can be surfaced before the output is trusted.
For difficult work, a second model or reviewer may be used as comparison assistance. It is additional review support, not an infallible judge.
What is acceptance evaluation?
Acceptance evaluation checks whether the output satisfies the work that was actually requested. It should inspect artifacts, evidence and criteria rather than trusting a caller’s claim that the task is complete.
This can include criterion-by-criterion evaluation, artifact review, missing evidence detection and unresolved gap reporting.
Is Graphode only for coding?
No. Software architecture and AI-assisted implementation are major use cases, but Graphode is broader. It can model APIs, data structures, workflows, organizations, compliance logic, product blueprints, operational systems and custom domains when the work needs structured truth, validation and proof.
What is MCP in Graphode?
MCP is the machine-facing tool surface that lets compatible assistants work with Graphode without turning the public app into the only place where work happens.
Through MCP-style tools, an assistant can fetch scoped context, report progress, create checkpoints, register artifacts, raise gaps, evaluate acceptance and resume work.
Which MCP tools are planned?
The public MCP catalog groups tools by user value: session and scope, task execution loop, GraphoTree read/context, draft editing, review/publish, artifacts and acceptance.
Exact availability may vary during early access. Tool names are presented as product direction unless the implementation has verified owner routes, schemas, claims and audit behavior.