The Foundational Unified Record Ecosystem
Your AI Forgets. Your Institution Cannot.
Long AI sessions erode constraints and compliance. FUR preserves institutional memory to keep initial decisions enforceable.
No data leaves your systems.
Developer tools free · Historian institutional
Liu et al. · Shi et al. · Min et al.
How FUR Works
From scattered sessions to structured institutional memory.
FUR consolidates separate AI conversations into structured, auditable records. Those records are stored in a persistent journal and re-introduced at the start of any new session — ensuring prior decisions, constraints, and context are never silently discarded.
Source conversations
Session A
Claude
Session B
ChatGPT
Session C
Gemini
FUR Conversation
Multiple chats become one structured, reusable conversation object.
FUR Journal
A journal can hold many FUR conversations, not just one chat history.
Re-introduced context
New AI Session
context restored from FUR
The Risk
AI Systems Forget. That Is a Compliance Problem.
It is not a malfunction. It is how the technology works. And in regulated, high-stakes environments, the consequences are institutional.
Critical Instructions Are Buried and Ignored
AI systems do not read a long session the way a human analyst reads a brief. Attention is distributed unevenly — information in the middle of an extended session receives a fraction of the weight given to the opening and closing. A constraint established at the start of an analysis is often invisible to the system by the time a conclusion is drawn.
The Research
Liu et al., Stanford (2023) — Documented systematic performance degradation when relevant information is positioned in the middle of extended AI sessions.
Exploratory Work Degrades Final Reasoning
Productive analysis involves exploration — hypotheses tested and discarded, directions pursued and abandoned. But every exploratory exchange that remains in the session record actively degrades the quality of subsequent reasoning. The AI system cannot distinguish between the inquiry and the conclusion.
The Research
Shi et al., Google DeepMind (2023) — Irrelevant session content degrades AI reasoning performance by up to 40% across major model families.
Established Requirements Are Silently Abandoned
A regulatory restriction introduced early in a session. A risk parameter established during initial scoping. A compliance boundary defined at the outset. By the thirtieth exchange, the AI system has effectively set these aside — not through error, but through the ordinary mechanics of how context is processed. The constraints have not been violated. They have been forgotten.
In legal, regulatory, or financial analysis, this is an institutional liability.
Observed In Production
Documented in extended production sessions across Claude, GPT-4, and Gemini in professional deployment contexts.
The Consequences Are Operational, Not Theoretical.
Legal Practice
Clauses missed. Jurisdictional requirements forgotten. Constraints that were established explicitly at the outset absent from the final work product.
Financial Analysis
Risk parameters that drift. Compliance thresholds that disappear. Consequential decisions produced by AI that no longer reflects the constraints under which it was engaged.
Institutional Research
Methodology that cannot be reproduced. Analytical parameters lost mid-investigation. Months of structured inquiry invalidated by undocumented session drift.
Technical Teams
Architectural decisions abandoned mid-session. Requirements that were clearly specified at the start absent from the final output.
How It Works
Persistent Constraints. Auditable Sessions. Reliable Output.
FUR treats AI sessions as structured institutional records — not disposable chat logs. Your team determines what persists, what is carried forward, and what is set aside. The AI operates only on what you have explicitly authorized.
Capture
Import session records from any AI platform — Claude, ChatGPT, or others. FUR accepts standard export formats and converts them into structured, referenceable records.
Curate
Designate which exchanges establish binding constraints, which represent exploratory inquiry, and which can be safely excluded from future sessions. FUR preserves your institutional decisions, not the full session noise.
Reintroduce
Before any new session begins, FUR reintroduces the curated record. Constraints remain active. Context is restored. The AI operates with the full benefit of prior institutional work.
Without FUR
The AI processes every exchange — productive analysis and discarded tangents alike
Attention is distributed indiscriminately across signal and noise
Established constraints erode as the session extends
By the twentieth exchange, original requirements may no longer be active
Output reliability degrades with session length
With FUR
The AI operates only on what your team has designated as relevant
Established constraints remain active regardless of session length
Exploratory work is preserved for reference but excluded from reasoning context
Session 100 reflects the same institutional constraints as session 1
Output is consistent, reproducible, and defensible
Why This Produces More Reliable Results
Not magic. AI systems have defined limits on the information they process in any given session. When that space is occupied by exploratory tangents and discarded directions, the constraints your institution established are crowded out. FUR reclaims that space for what matters.
Research-validated. Peer-reviewed research from Stanford and Google DeepMind confirms that curated context produces substantially better AI reasoning. FUR operationalizes that finding — systematically, reproducibly, and without requiring technical expertise from your team.
The Products
One Foundation. Two Interfaces.
The FUR Engine is the underlying system for structured AI context management. Historian is the institutional interface built on top of it — designed for professionals who require results, not a technical learning curve.
FUR Engine
Foundation · Open Source · MITThe core system for structured AI session management. Maintains version-controlled session records, supports branching analysis paths, and provides selective context reintroduction. The basis for everything else in the ecosystem.
FUR CLI
v1.0.0 · Open Source · MITFor technical teams and developers who need to integrate structured AI context management into existing systems and workflows. Command-line native, scriptable, and fully transparent.
Historian
Institutional · CommercialA desktop application for professionals and institutions who need structured AI session management without a technical interface. Local, private, and fully under your control.
Why Two Products?
A central bank analyst and a software engineer solve the same underlying problem through entirely different workflows. Historian is built for professionals who should not need to learn a command-line interface to protect institutional memory. FUR CLI is built for the technical teams who support them.
FUR CLI — For Technical Teams
Integration-ready. No abstraction layer.
Historian — For Institutions
Professional interface. No technical requirement.
Why Historian
What Standard Tools Cannot Provide
General-purpose AI platforms and note-taking tools were not designed for the accountability requirements of regulated professional environments. Historian was.
vs. ChatGPT / Claude directly
vs. Obsidian
vs. Notion
Capabilities
| Feature | Historian | ChatGPT | Notion | Obsidian |
|---|---|---|---|---|
| Local data control | ✓ | — | — | ✓ |
| Structured AI session management | ✓ | — | ✓ | — |
| Full session length support | ✓ | — | ✓ | — |
| Source attribution and auditability | ✓ | — | ✓ | — |
| Open, portable format | ✓ | — | ✓ | ✓ |
| Exchange-level constraint curation | ✓ | — | ✓ | — |
| No vendor dependency | ✓ | — | — | ✓ |
Legal Counsel
Every regulatory constraint and jurisdictional requirement remains active through the full analysis. No clause missed because the AI lost the thread.
Financial Analysts
Risk parameters and compliance thresholds established at the outset remain operative at the conclusion. Assumptions do not drift. Outputs are defensible.
Institutional Researchers
Methodology is preserved and reintroduced at each session. Analytical parameters remain consistent across weeks of investigation. Findings are reproducible.
Your Constraints Should Outlast Your Session.
Whether your institution uses AI for legal analysis, financial modeling, policy research, or regulatory review, FUR ensures the requirements you establish at the start remain active at the end.
Historian is available for institutional deployment. FUR CLI is available open source for technical integration.
For Technical Teams
FUR CLI is open source and available for integration into existing systems and workflows.
Explore FUR CLI →For Institutions
Historian is currently in restricted institutional deployment. Contact us to discuss your organization's requirements.
Request Access →