Giraffe: A Readable Project Record for the Work You Do With AI
AI conversations help us think—but they rarely show, at a glance, where a project stands today. Giraffe turns the conclusions that matter into a clear, living record: current facts, decisions, open questions, and the history behind every change.
AI conversations help us think—but they rarely show, at a glance, where a project stands today.
Giraffe turns the conclusions that matter into a clear, living record: current facts, decisions, open questions, and the history behind every change.
What is Giraffe?
Giraffe is a project memory platform for work done with AI assistants.
You continue researching, comparing, planning, and deciding in the assistant you already use. When something becomes worth keeping, you explicitly ask to save it to a Giraffe Project—or add it directly in Giraffe. The result is organized into a record that can be read without reopening the original conversation.
A Project can contain:
- Current: the facts, values, dates, requirements, and status that apply now;
- Decisions: the choices that were made, including decisions later reversed;
- Open questions: the issues that still need an answer;
- Notes: useful context that does not belong in the other categories;
- Documents: longer outputs created from the project;
- History and recent changes: prior values and the sequence of updates;
- Sources: information about where an update came from, when available.
This structure makes Giraffe different from a folder of transcripts. A transcript preserves the route through a discussion. A maintained Project presents the current destination while retaining the route when it matters.
The problem: a conversation is not a project record
Imagine planning a product launch with an AI assistant.
In the first week, the target date is November 15. A later conversation moves it to December 1. One feature is dropped. The audience changes. A launch channel that looked promising is rejected after new research. Two questions remain unresolved.
All of that may exist in chat history, but history alone does not tell you which statements still apply. The first date is just as searchable as the second. The rejected feature can look current when separated from the message that replaced it. The final decision may be buried below several plausible alternatives.
Giraffe approaches the same project as maintained state:
| Project element | Example |
|---|---|
| Current | Launch date: December 1 |
| Current | Scope: core workflow only |
| Decision | Defer the reporting feature until after launch |
| Open question | Is partner approval required before announcement? |
| History | Launch date changed from November 15 to December 1 |
The value is not simply that information has been saved. It is that the saved information says what kind of thing it is—and whether it is still current.
How Giraffe works
1. Work in the AI assistant you already use
Giraffe is designed to sit beside existing AI workflows. Its official setup documentation currently covers Claude, ChatGPT, and Gemini, with different connection requirements for each service.
You can also use Giraffe directly in the browser without connecting an assistant. That matters because the product is useful even for someone who uses only one AI—or prefers to maintain a Project manually.
2. Save only what matters
Nothing is automatically copied into Giraffe. The user chooses what to keep.
After reaching a useful conclusion, a person might say:
Save what we concluded to my launch project.
The assistant can file the relevant information into the Project’s existing structure. A confirmed date belongs in Current. A settled choice belongs in Decisions. An unresolved dependency belongs in Open questions.
This explicit model is an important product boundary. Giraffe is not an automatic backup of every AI conversation, and it does not claim to resolve every change on the user’s behalf.
3. Open the Project and read where things stand
The Project itself is meant to be readable. A user can open it before a meeting, return after two weeks away, or look up one decision without prompting an assistant to reconstruct the answer.
Search works across Projects and across structured types such as current facts, decisions, documents, and past values. That shifts retrieval from “Which chat contained the sentence?” to “What did we decide?”
4. Update the same record as reality changes
Projects evolve. Budgets move, dates slip, requirements narrow, and decisions are reversed.
Giraffe keeps one live value for what is true now while preserving the value it replaced in history. A reversed decision can remain visible with its reason. This creates a record that is current without pretending the past never happened.
Three practical benefits
A clear current state
The strongest benefit is simple: Giraffe separates current information from the discussion that produced it.
For a renovation, that could mean the active budget, selected contractor, agreed materials, start date, and remaining question about permits. For a job search, it could mean current salary criteria, target roles, rejected paths, and unanswered questions for the next interview.
The user reads the Project as a snapshot, not an archaeological dig through old messages.
Decisions with context and history
A decision is more useful when it stays connected to what it replaced and why it changed.
Giraffe retains old values rather than silently overwriting them. Its decision model also supports marking a decision as reversed. That is useful for long-running work, where remembering the reasoning behind a previous direction can prevent the same debate from starting again.
Continuity without depending on one conversation
The conclusions live in the Giraffe Project rather than solely inside one transcript. A user can read the record directly, copy it for use elsewhere, or ask a connected assistant to read it before continuing.
This does not mean assistants synchronize their chats with one another. They do not. The continuity comes from reading and updating the same explicitly maintained Project state.
Where Giraffe is especially useful
Giraffe fits work that changes over time and will need to be revisited.
Preparing for a client meeting
A consultant can keep the client’s current situation, prior decisions, assumptions, and open questions in one Project. Before the next meeting, the useful preparation is already separated from weeks of exploratory conversation.
Planning a launch
Dates, scope, channels, dependencies, and positioning often change independently. Giraffe gives each current value a clear place while preserving the prior state.
Conducting long-term research
Not every research project ends in one dramatic decision. Some accumulate conclusions over months. A maintained Project keeps the useful findings accessible without requiring the researcher to remember the original query or conversation.
Comparing a major purchase
Requirements can change as research progresses. The budget may move, the shortlist may narrow, and a rejected option may become relevant again. Keeping current criteria and earlier reasoning together makes the comparison easier to revisit.
Managing a personal project
Travel plans, home renovations, career decisions, and event planning all generate changing facts and unresolved questions. Giraffe Personal is designed for an individual maintaining that evolving record, not as a team knowledge-management promise.
Giraffe versus chat history and native AI memory
These tools serve different purposes and can coexist.
| Tool | Primary job | Best question |
|---|---|---|
| Chat history | Preserve the conversation | “Where did we discuss this?” |
| Native AI memory | Help an assistant personalize future responses | “What should the assistant remember about me?” |
| Giraffe Project | Maintain the state of a concrete undertaking | “What is true now, what did we decide, and what is still open?” |
Chat history is valuable evidence of how the thinking unfolded. Native memory can help make later responses more relevant. Giraffe is intended to make the record itself readable, including the present state and the values or decisions it replaced.
What Giraffe does not do
A precise description is more useful than an inflated one.
Giraffe does not automatically record every conversation. Nothing reaches a Project unless the user asks an assistant to save it or enters it directly. That protects the product’s selective model, but it also creates a real limitation: users must build the habit of saving meaningful conclusions.
Giraffe’s official site also states that existing conversation archives cannot currently be imported. A new account therefore starts with the next useful item the user saves, rather than reconstructing an entire backlog automatically.
Connections have platform-specific requirements. As of the official documentation reviewed for this article, ChatGPT setup requires a paid plan on the web with Developer mode enabled and does not work inside a ChatGPT Project; Gemini uses Gemini Spark; Claude has its own connector setup. These conditions can change as the underlying platforms evolve, so users should check Giraffe’s current setup guides before connecting an assistant.
Export and ownership of the record
Giraffe states that Projects can be exported as Markdown and JSON, with documents, old values, and dates included. That makes the stored record portable and inspectable outside the interface.
The product is currently presented as free, with no payment card required. As with any live software product, availability, platform requirements, and pricing should be checked on the official site before relying on them.
A useful first Project
The best first Project is not the biggest. It is one you expect to reopen.
Start with a concrete undertaking—a launch, purchase, trip, renovation, research topic, or career decision—and save only four things:
- one fact that is currently true;
- one decision already made;
- one open question;
- one note worth retaining.
Open the Project directly. If it already tells you where the undertaking stands, the structure is doing its job. When one fact changes, update the same item and inspect the history. That small cycle demonstrates the central value of Giraffe more clearly than importing a large quantity of undifferentiated information would.
Frequently asked questions
Does Giraffe automatically save AI conversations?
No. Information is saved only when the user explicitly asks a connected assistant to save it or enters it directly in Giraffe.
Is Giraffe another AI chatbot?
No. Giraffe is a readable project record for the conclusions produced through AI-assisted work. The thinking can continue in the user’s existing assistant.
Can Giraffe show what changed?
Yes. The product keeps current values while retaining replaced values in history. Decisions can also remain visible when reversed.
Does Giraffe require multiple AI assistants?
No. It can be useful with one assistant or with no connected assistant at all. Cross-assistant continuity is an option, not the core requirement.
Can Giraffe import old conversations?
Not currently, according to the official product site. Users begin with the next information they choose to save.
Can Projects be exported?
Yes. Giraffe states that content can be exported as Markdown and JSON.
How much does Giraffe cost?
The official site currently presents Giraffe as free, with no card required. Check the product site for the latest terms.
The bottom line
The useful outcome of AI work is rarely the full conversation. It is the small set of facts that remain true, the decisions that now govern the work, the questions that still block progress, and the history that explains how the project arrived there.
Giraffe gives those conclusions a stable, readable home. Its value comes from disciplined selectivity: the user decides what matters, the Project shows what applies now, and earlier states remain available when context is needed.
Explore how Giraffe works or review its current use cases.
Start with Giraffe: https://withgiraffe.com/
