Clear answers before you start.

Fileaify brings together project documentation, its maintenance rules, service references and permissions granted to people and agents. You can start a fresh conversation in the same tool or choose another agent: an authorized session retrieves published context and can propose updates. Project knowledge stays independent of a conversation history, a tool and a machine. Development continues in your usual tools; Fileaify does not replace your repository or execution environment.

Yes. This is a core use case: stop depending on an ever-growing conversation that mixes every project topic. Before leaving it, the agent proposes decisions, results and useful information to keep. Once published, those facts can be read by a fresh authorized session, even in the same tool. Secrets remain in the optional vault with separate permissions. Fileaify does not control your tool’s context window or compaction and guarantees no specific speed or token savings.

Yes. Two uses are equally important: starting a fresh conversation without remaining tied to a long history, and freely choosing the agent tool you work with. In either case, an authorized session can retrieve published context without access to the previous conversation. Your personal workspace is created automatically: no team or group setup is required. Permissions separate your rights from those granted to agents.Working alone, it is easy to move fast and document little. The template provides a method you might not have written yourself, and Fileaify gives it to each agent: what to document, where it belongs and when to update it. This structure is useful even with a single project and the same tool.

Yes. Reading authorized projects, retrieving the documentation contract, proposing changes and configuring publication policy are available on Free. The built-in assistant is separate and is not required for these contributions. Free includes one workspace, one active project, five services and five secrets. Your external agent tool’s costs remain separate.

Create an account and a project from a template. Give your agent https://fileaify.com/agent and the project name. It prepares its connection and presents an approval link: you choose the project, permissions and duration. The agent then makes the technical calls, reads the rules and documentation, and prepares updates related to its work. You can keep this routine in your tool’s instructions. No custom API client or vault secret is needed to get started.

A single project developed over several months is enough: architecture decisions, deployments, interface work and business tasks accumulate. Fileaify helps when you want a fresh conversation without reconstructing that information, or the freedom to choose another agent tool. It also supports resuming after a break and handing work over. The gain is in the effort spent finding context, repeating explanations and maintaining the documentation setup: a well-kept AGENTS.md covers part of it, and Fileaify brings the rest (permissions, vault, review, history) in one framework.

Git versions code and can also hold documentation maintained by agents: a well-kept AGENTS.md already does part of the work, and Fileaify connects to it instead of replacing it. What Fileaify adds: permissions per agent, with a duration chosen by the person who accepts; a vault whose keys the agent uses without ever receiving the value; review of contributions before publication; the history of decisions; an executor to act on your servers without holding a key; and a maintained context shared across your tools and sessions. You save time (no context to rebuild at each session, no permissions to improvise project by project), gain security (no keys in conversations, time-bounded access) and organization (a common method, stable conventions, an up-to-date documentation state).

No. Code stays in its repository, and you do not need to maintain two copies of documentation. Choose where each document is authoritative: some files can remain in Git and shared context in Fileaify. Fileaify does not automatically synchronize with GitHub. Consistency between copies only needs managing if you choose to duplicate a document in both places.

One repository, stable conventions, no important decisions outside the code and no credentials to delegate: when your project meets these conditions, a well-maintained AGENTS.md or CLAUDE.md can suffice. CLAUDE.md is Claude Code’s instruction file. Fileaify plugs into these files to provide a managed system: a reusable documentation method, published context, services and controlled access. An instruction pointing to the project lets an authorized agent retrieve this context in a fresh conversation or another tool. You retain decisions and approvals; the agent follows the protocol and proposes updates, without a custom integration for you to develop and maintain.

Local memory or an MCP server can cover part of the need and offer other benefits, including local control. Capabilities depend on the tool. Compare the full workflow: access across tools, documentation method, versions, publication, services, optional secrets and revocable permissions. Combining these functions can require integrations and maintenance; Fileaify provides them within one hosted service.

An agent must be able to read the protocol, make the required HTTPS calls, send JSON and privately use a Bearer token. A terminal-equipped agent can use curl, so you do not need to develop a custom client. Tools listed on the homepage are not individually certified native connectors: operation depends on their mode, network capabilities and permissions. A remote MCP connector is also available at https://fileaify.com/mcp with OAuth: the client handles calls and opens Fileaify for access selection. Available features depend on your provider account.

It can read public pages, including the protocol and this FAQ. Pairing requires POST requests, followed by authenticated calls to read private projects. A search browser restricted to public URLs is therefore insufficient. Do not work around this by putting a token in a URL or making private documents public. If the chat supports MCP connectors, add https://fileaify.com/mcp with OAuth: the connector handles authenticated calls even when the chat’s search browser only supports GET. The browser alone remains insufficient.

Your external agent can do this as it works: it reads project rules, classifies facts and proposes changes with rationale and supporting evidence. The built-in assistant is not a required intermediary. Fileaify does not monitor your terminal or every conversation: facts that are not submitted are not known. A persistent instruction in your tool helps the agent check necessary updates at the end of each task.

You choose the project and its template, authorize your agent and decide whether its updates require your review. The agent makes API calls, reads the rules and prepares documentation contributions. Fileaify operates the hosted service and enforces permissions: you do not need to program an API client, install a Fileaify server or maintain that assembly. A valid token serves multiple calls until it expires or is revoked. Your involvement remains necessary for project decisions, additional access and the reviews you choose.

It defines expected documents, their purposes, included and excluded content, update triggers, instructions and Markdown formats. Start from a template, a blank project or add your own documents. Administrators edit project rules; agents read them through a versioned contract. Updating a reusable template does not automatically replace the rules of existing projects.

No. Your workspace’s reusable documentation templates define documents, purposes, formats and update triggers. Use them for new projects and adapt their rules as needed. Each authorized agent reads the project’s effective method through its contract, so you do not need to copy it into every conversation. A template update applies to future projects; it does not automatically change existing projects. Template choice and project-specific adjustments remain under your control.

Human review is the default. An administrator can allow automatic publication for selected agents or all agents with write permission, per project and on every plan. Permission, contract and revision checks still apply. A pending proposal is not published; changing policy does not retroactively approve existing pending proposals.

A pending transaction locks its affected documents until approval or rejection. Other agents can still read the project, but edits to a locked document are rejected. An outdated base revision also causes a conflict. The agent must reread the context and prepare a coherent proposal; there is no automatic text merge.

With or without Fileaify, instructions do not guarantee that an agent follows them. With Fileaify, every documentation contribution includes its rationale and supporting evidence: you review it, or use the automatic publication policy you chose. Fileaify flags documents that are still empty or retain their initial content, and unresolved key references. These gaps become visible; this does not detect every omission or factual error.

No. A token is delivered once during pairing and then reused for API calls until it expires, is suspended or is revoked. Pairing is not required for every request. Store the token in the tool’s private credential storage. Usage limits on secret delivery are separate from token reuse.

A workspace owner or administrator checks the request and chooses the project, capabilities and duration. The agent does not inherit that person’s role. Pairing grants no secret value. Fileaify controls its own projects and resources; it does not restrict GitHub permissions or local access already held by the agent’s tool. Additional requests remain separate, and a document reference never creates a permission.

The app’s Pending requests drawer groups the decisions available to the user by project. Pairing requests are presented through their matching link or code. The PWA remote also supports permission management on a phone. Notifications depend on browser support and granted permissions; dismissing a notification is not a decision.

Yes, with the required administrative permissions. Suspension blocks access and can be lifted; revocation invalidates the agent. Grants can also expire or be limited to specific projects and resources. These actions do not retrieve documents or secrets already delivered to the agent.

The executor is a small server dedicated to your workspace that can use your declared SSH keys on your behalf, so agents can act on your systems without ever receiving the key. It is available now. By plan: Solo on demand for 1 hour per 6-hour window, Pro on demand for up to 12 hours per day, Premium always on. The IP address changes at every start; Pro and Premium can turn on the “Fixed IP” option in My account, applied at the next start. Free does not include it. It only connects out, to Fileaify and then to the targets you declare. Its single inbound door is a range of SSH ports (one port drawn at each start, key authentication only) used when an agent you authorized opens a direct session to its own project folder; a human decides how long that permission lasts.

No. You can use the documentation method and project memory without storing any secret. Later, you can add useful resources with dedicated keys and limited grants. Credentials kept in your own tools remain managed there; Fileaify cannot apply its controls to a key shared outside its vault.

A service identifies an account or provider used by the project. A connection describes a target or use, such as an environment and URL. A secret holds an encrypted value and its versions. Project bindings and permissions are explicit. A vault:// reference lets documentation identify a key without containing its value.

Objects are stored in a private S3 bucket with AWS KMS encryption; public connections use HTTPS. Permissions are checked server-side and every access to a secret leaves an audit trace. Encryption protects objects at rest; the server decrypts for authorized operations. The security notice describes the measures.

Yes, for configured HTTPS brokering: an administrator defines an exact destination and allowed methods, then grants separate usage permission. Fileaify adds the credential to that request. This does not cover every tool or provide an SSH terminal. Exporting a credential to an agent requires separate permission. Revoking delivery does not disable a previously received copy: rotate the credential with its provider.

Yes, names and usage only. Any agent allowed to read a project can list its vault references (for example vault://brevo/key) and what each key is used for in that project, never the values. To use or export a key, it asks for a separate permission on that specific key, which a human approves.

By default it never can: export is off. It must be enabled at three levels, in My account, in the workspace settings (owner only) and in each project concerned, and every delivery also needs a human approval limited in time and uses. The value is then sent to the agent, which holds it: it may enter the model context at its provider. Preferred: with an executor the agent uses the key by reference (vault://…) and never sees it. Without an executor, for example on the free plan, it can write a small local routine that calls the Fileaify API with its token, writes the answer to a file and reads that file without ever printing it.

It provides optional assistance in workspaces whose plan includes AI. The global assistant reads authorized projects without editing documents. The project assistant works within that project, using Plan, Ask or Autonomous mode and its review setting. It also helps design documentation templates. Your external development agents continue working without it.

Built-in AI features use Mistral and send the context needed for the feature: the question, limited history, authorized documents, rules and attachments. The vault is not loaded into that context: keep your secrets in the vault rather than in documents or messages. Data sent to Mistral is not used to train its models. External agents use their own providers and policies. Their access to Fileaify does not automatically invoke the Mistral assistant.

Three levels, each with its own terms: Fileaify storage, Mistral processing for built-in features, and your external agent’s provider. Data sent to Mistral is not used to train its models. The security notice and the privacy policy detail what is transmitted and retained.

AI reviews are scheduled and explicitly enabled by an authorized person. Configurable criteria help examine documentation, gaps or inconsistencies. Reviews produce reports and do not automatically apply corrections. They are not a mandatory LLM gate for every external-agent contribution.

Free includes documentation and external agents within its capacity limits. Solo increases capacity without the built-in assistant; Pro and Premium include built-in AI. Current offers and availability are shown on Pricing. Paid subscriptions are confirmed in Stripe, monthly or yearly, and renew automatically. Cancellation from My account takes effect at the end of the paid period. Plans manually granted to testers never trigger a charge.

No. Each workspace inherits its creator’s plan. Project limits and Free service/secret limits are counted across that creator’s workspaces. Members are counted uniquely across those workspaces. Free and Solo do not allow additional invited members. Permissions still belong to individual workspaces and projects; available capacity does not grant access.

A preview identifies projects beyond the new capacity. The least recently updated projects become inactive and read-only; their documentation remains downloadable. Deleting another project does not let you freely switch a locked project back to active. Increasing capacity can reactivate projects under the plan rules.

No. The built-in assistant quota is separate from your external tools’ usage. Defaults are 10 calls per five-hour window on Pro and 50 on Premium; administrators can configure them, and they are shared according to workspace creator. The API still has its own controls and technical limits: no assistant quota does not mean unlimited API requests.

Yes. The document ZIP contains published documents and their contract, without vault values. A separate complete encrypted export is available in My account to the workspace owner after recent sign-in and explicit authorization. It includes retained document and vault versions and files, rules, proposals, services, permissions and audit records. Other members’ private chats, deleted or cross-workspace chats and login credentials are excluded. A separately saved key and offline tool recover a Markdown/JSON/files ZIP without Fileaify or AWS access. Beyond the 64 MiB direct limit, free assisted restitution is available within thirty days of an authenticated request. Your Git repositories stay with their host.

The application runs on an OVHcloud VPS. S3/KMS storage and Cognito authentication use AWS eu-west-1 in Ireland. The privacy notice presents the other providers and their role.

Assess verifiable commitments: free data portability, an encrypted vault export recoverable offline, daily encrypted metadata backups with restore checks and an off-VPS copy, document and secret objects in versioned S3 encrypted with KMS, and at least sixty days’ notice for a planned shutdown. The published policy defines the scope. These facts let you assess continuity independently of the publisher’s size.

Archiving keeps the data. For an erasure, send a request to support: the operator deletes the objects and versions concerned and confirms it to you. The privacy notice describes retention.

The interface is available in English and French. A project’s documentation language is separate: it guides templates and writing instructions without automatically translating existing documents. Planned additional languages are not presented as available until their translations are delivered.