Own your context
Keep project memory, handoff packs, decisions, and proof in a HeapFile-controlled layer instead of trapping them inside one chat or vendor.
Private. Portable. Provable.
Own your context.
Lease it to AI.
HeapFile makes AI work continuous, private, portable, and provable. It keeps the source of context under your control, sends only the right slice to eligible local or authorized providers, and leaves a receipt you can review.
Context stays yoursLocal-first planningSavings only when measured
HeapFile UI manages accounts, billing, downloads, and support. HeapFile AI is the installed local workspace/runtime that carries the work.
Local plan before provider calls
Discover. Route. Prove.
Capability discovered
ObserveOnly observed local or authorized routes become eligible for the task.
Context leased
RouteThe smallest useful packet carries sources, limits, and approval boundaries.
Proof receipt
ProveMeasured facts, estimates, self-reports, and unknowns stay separate.
Illustrative workflow, not live customer activity.
Let your AI learn it for you
Don’t read the docs.
Point your AI at HeapFile.
heapfile.ai is built to be read by agents. Paste the prompt below into your assistant and it will explain what HeapFile does, which features are free, and how to set it up as an MCP server for your workflow — so you don’t have to read the docs yourself.
Read https://heapfile.ai/llms-full.txt and https://heapfile.ai/features.json, then explain what HeapFile does, which features are free, and how to set it up as an MCP server in my AI client for my workflow.Prefer to point your agent at a specific resource? heapfile.ai publishes machine-readable references built for agents:
- llms-full.txt — complete product description, including every free feature.
- features.json — machine-readable Free vs Premium feature manifest.
- llms.txt — concise, agent-friendly index of HeapFile.
- start.html — a short “teach your AI” walkthrough.
The model is rented. The context is yours.
Your AI should work from context you control.
AI gets useful when it has context. That context becomes operating knowledge: files, decisions, workflows, customer nuance, and proof of what happened. HeapFile keeps that knowledge portable and leased to the right AI for the right task.
Discover what is actually available
HeapFile separates supported contracts from observed capability on your machine, so unavailable providers stay marked unknown.
Route with boundaries
Use eligible local or authorized providers with explicit cost, locality, fallback, and permission boundaries before work runs.
Let safe work continue
Deterministic local workers can queue, retry, and refresh bounded tasks within policy. Consequential actions still need the right approval path.
Prove what happened
Review context used, route selected, output produced, proof collected, and what stayed blocked or unknown.
Scale when continuity matters
Premium expands connector breadth, retained context, background refresh depth, and exportable proof/reporting for more serious workflows.
One local-first runtime flow
Find the path.
Pack the context.
Leave proof.
Planning, capability negotiation, context packing, queueing, and proof collection run locally without hosted model calls. Later provider execution is separately authorized, measured, and policy-gated.
- 1Discover available AI
HeapFile checks observed local adapters, authorized hosted lanes, files, connectors, and tools. A supported contract is not treated as live access.
- 2Build the smallest useful context
The runtime packs sourced evidence, decisions, constraints, allowed actions, blocked actions, and exact local measurements.
- 3Route through an eligible provider
HeapFile chooses a local or authorized route only when the task requirements match observed capability and declared fallback semantics.
- 4Continue deterministic work in the background
Registered local handlers can run in a leased, retryable, resource-bounded queue. Outbound sends, file delivery, purchases, permission changes, and model calls stay separately gated.
- 5Record a reviewable proof receipt
The Value Ledger records what was observed, what was estimated, what came from a provider self-report, and what remained unknown.
- 6Show measured savings only when comparable
HeapFile claims savings only when the HeapFile run and a baseline share the same comparison key and comparable metrics.
For the way you work
A shared brain layer.
Many useful workflows.
Everyday AI users
Stop explaining the same task.
Carry the goal, relevant context, and next step into the next AI session without giving one vendor your whole history.
YouTubers & creators
Keep research and scripts moving.
Save channel context, video research, source trails, and script decisions so future searches and drafts start informed, with review before publishing.
Small business marketing
Keep moving when one channel is blocked.
Save campaign decisions and next steps across AI sessions. See a practical handoff example and what is still in development for Marketing Assist.
Explore the marketing workflowOperators & support teams
Make handoffs less painful.
Give the next session the case history, constraints, and proof trail instead of repeating discovery.
Engineering & AI platform teams
Route work without a custom framework.
Use provider-neutral work packets, local capability checks, bounded workers, and evidence-rich handoffs across compatible agents.
Managers & finance leaders
See value without fuzzy math.
Separate measured savings from estimates, provider self-reports, and unknowns before using the numbers for decisions.
Security & governance teams
Know what AI received.
Review provider boundaries, authorization scope, task policy, proof receipts, and export needs before expanding access.
Explore trust & proof →The HeapFile suite
One product.
Two clear roles.
HeapFile UI is the account and subscription surface. HeapFile AI is the installed local runtime. Your chosen model or tool still supplies its own access and cost.
Account, billing, downloads, support
Use heapfile.com to create an account, manage live PayPal billing, request support, and access current account-gated release downloads.
Installed local workspace/runtime
The desktop control plane shows the Capability Router, Worker Mesh, Provider Boundary, Value Ledger, queue/proof state, Autonomy Ladder, compatibility freshness, and local proof cycle actions.
Observed or authorized routes
HeapFile can describe many provider families, but it selects only what is actually configured and allowed. AI-provider subscriptions, API usage, storage, and connector costs are billed separately.
Connectors and compatibility
Connect the context you already control.
HeapFile brings selected local files, repositories, work systems, cloud platforms, and AI providers into a local-first context layer through user-authorized connectors. Every connector is labeled by actual readiness, required authorization, permission level, and data route.
Before connecting a project
Choose one source and check four things: whether its connector is ready in your release, who must authorize access, whether the scope is read-only or permits writes, and where the data travels. Start with an invented project or a small non-sensitive source. A ready connector does not grant permission to publish, deploy, or change an external service.
Product, account, and support
Use this site to evaluate HeapFile, create an account, manage plans, reach support, review trust materials, and access current downloads.
Setup and connector status
Use heapfile.ai for installation, setup, technical workflows, connector compatibility, permissions, and machine-readable capability information.
Names are compatibility labels
Provider names identify compatible or planned systems. They do not imply endorsement, sponsorship, partnership, or authorization. Availability depends on the HeapFile release, provider terms, account plan, organization policy, user or administrator authorization, and configuration.
Local-first, not local-only: HeapFile is local-first, not local-only. Local sources and eligible local model routes can remain on the device. When a user deliberately connects a cloud service, that connector contacts its provider through the authorized route and permission scope shown during setup.
Free vs paid
Starter is live.
Premium expands the operating layer.
Starter is useful without a paid HeapFile subscription. Premium is for broader connector reach, retained continuity, background refresh, and proof/reporting workflows.
Available now: create an account, start with Starter, upgrade through live PayPal billing, and use the account portal for current HeapFile AI downloads and support.
Starter · Free
$0/ month
For an individual starting with the local-first core.
- Local planning and context packs
- Actually available local adapters
- Device-local proof receipts and handoffs
- Human approval gates for important actions
Create an account and start with the local-first core. No HeapFile paid subscription required.
Start freePaid tier
Premium
$19/ user / month
For individuals running more connected workflows with supported setup.
- Everything in Starter
- Authorized connector breadth during onboarding
- Retained and portable context availability
- Deeper background refresh and proof/report export
The live monthly Premium plan. Sign in to subscribe, review status, download current builds, or manage cancellation.
Manage live billing Live PayPal billing · $19 per user/month planEnterprise
Let’s talk
For organizations evaluating governance and deployment needs.
- Deployment and integration planning
- Role-based context access and policy review
- Retention, export, BYOC, and rollout scoping
Custom scope by agreement. Enterprise controls are not included simply by selecting a plan.
Discuss your requirementsWhat your membership pays for: HeapFile functionality and the agreed onboarding scope. It does not include Claude, ChatGPT, Codex, Ollama hosting, other provider subscriptions, API credits, or unlimited AI usage. Those costs stay with your chosen provider.
Proof before claims
Savings should be earned
by comparable evidence.
HeapFile can show value in different evidence states. Only measured, comparable baselines should become claimable savings.
Claimable when comparable
HeapFile and the baseline share the same comparison key and comparable metrics.
Useful forecast
Planning calculators and projected value stay visible but separate from measured savings.
Source-labeled evidence
Provider or upstream reports are shown with attribution and excluded from claimable totals.
No invented numbers
Missing baselines, unavailable usage, and unsupported routes remain unknown instead of turning into zero.
Local-first economics: planning, capability negotiation, context packing, queueing, and proof collection can run locally at negligible hosted-model cost. Provider calls, storage, connectors, subscription infrastructure, and your model/API accounts are separate costs.
Estimates are not guaranteed savings. Include only spending you can actually eliminate, including any measured reduction in provider usage. Avoid counting the same benefit twice.
Straight answers
Know what you’re
starting with.
What is HeapFile?
HeapFile is the continuity, privacy, portability, and proof layer for AI work. It helps carry relevant context, task progress, and reviewable outputs across compatible sessions and tools while keeping the context source under your control.
What is free, and why would I pay?
Starter is the $0 entry point for local planning, local adapters that are actually available, device-local context/proof, and handoff packs. The $19 per user/month Premium plan expands connector breadth, retained context availability, background refresh depth, and proof/report export as configured during onboarding.
Does HeapFile include an AI subscription?
No. Keep your own AI-provider account, local model, or API access. HeapFile membership does not include provider subscriptions, API credits, storage bills, or unlimited model usage.
Does it work with every model and tool?
No universal support is claimed. HeapFile can describe hosted OpenAI, Anthropic, Google, local Ollama-compatible and OpenAI-compatible endpoints, MCP, files, Git, and connectors, but a route becomes selectable only after it is observed, configured, authorized, and eligible for the task.
What is the difference between listed, authorized, and partnered connectors?
A listed connector means HeapFile has a compatibility record or planned route. An authorized connector means a user or administrator has completed setup and granted the required scope. A provider partnership means a separate verified relationship with that provider. Provider names on HeapFile pages do not imply endorsement, sponsorship, partnership, or authorization.
Can AI act without my approval?
Bounded local workers can run registered deterministic handlers inside policy. Outbound sends, file delivery, external writes, purchases, permission changes, and provider calls remain approval-gated unless a real provider explicitly supports a narrower authorized path.
Where do I manage or cancel billing?
Use the HeapFile account portal to review your billing status, start checkout, or request cancellation. PayPal also provides subscription management through your PayPal account.
Which desktop download options are listed?
The account page lists Windows, macOS, and Linux options. Sign in to review current build availability, system requirements, and access conditions for your account. Support can help if an expected artifact is not visible.
How can I evaluate AI memory privately?
Start with an invented project and compare what a fresh session recalls, how changed decisions are handled, and what context is shared. Read the evaluation guide before connecting sensitive work.
How do I send feedback?
Use the HeapFile feedback page for workflow observations or sanitized crash reports. Product access, billing, and download questions can go through HeapFile support.
Read all HeapFile FAQs · Why AI sessions lose context · Try a saved handoff · Keep marketing moving
Keep the context. Carry the proof.