Features
What you can do with docspi
Everything on this page works today. Each item also says who does it: you, on screen, or the AI agent working for you, through the API.
How to read this page
docspi has two kinds of user: you, in your browser, and the AI agents you work with, which reach it through an API. Every item below says which of the two performs it today, because it is not always both.
- In the app
- You do it yourself, on screen.
- Through the API
- Your agent — or any script — does it through the REST API or an MCP client. There is no screen for it yet.
- Signed-in request
- Available through the account API to someone signed in; agent tokens are refused, and there is no screen for it yet.
- Automatic
- docspi does it for you, with nothing to switch on.
- Set up with us
- Configured on your account by us rather than self-serve.
Where a paid plan is required it is marked. Every limit, and which plan unlocks which feature, is on the pricing page.
Bring scattered documents into one tree
docspi holds an index rather than your files: the structure, the names you choose, and the hash of each version it records. Building a tree copies nothing.
Create projects
In the appThrough the APIKeep each body of work in its own project and reach them all from your dashboard. How many an account may hold depends on the plan.
Build a virtual tree
In the appThrough the APIAdd folders and document entries, rename them, drag them to a different position or parent, and delete what you no longer need — all on screen. Through the API an agent can add and remove entries itself. The names in the tree are yours to choose; they do not have to match the file names underneath.
Keep a personal view
In the appCreate a personal view and arrange the same documents your own way in it, without changing what anyone else sees.
Search a project
Through the APISearch by name, by position in the tree, or by tag, and get the matching entries back. This is how an agent finds a document before working on it.
Notice when a document has moved or changed
In the appThrough the APIA client that can see your files reports the state of each mapped document. docspi records what no longer resolves, notifies the owner, and keeps a report you can read on the project's health screen — and when a file has moved you can repair the mapping by giving its new location.
Write with rules that are actually enforced
docspi checks and records document changes. The bytes stay in your own file — docspi returns the content it accepted and your client writes it.
Save a document version through docspi
Through the APIA save records the new version, its content hash, its place in the tree, and whether a person or an agent made it, and reports back whether the document was created or updated. docspi hands the accepted content back so your client writes the file itself, wherever it keeps it.
Set writing rules per project
In the appPro and aboveDefine the tone, the intended audience, the sections a document must contain, and a format template — per project, from the project's settings screen.
Have every save checked against those rules
Through the APIPro and aboveWith rules active, a save that misses a required section or breaks the format is rejected — with a score and the list of what failed — instead of quietly landing.
Ask for personal data to be masked as it is saved
Through the APIA save can request that personal data in the content be masked before the version is recorded.
Have every change recorded
Signed-in requestEvery save appends an entry by itself — sequence number, what happened, whether a person or an agent did it, and a content hash — so how a document came to be can be read back later by someone signed in to the account.
Hand a project to an AI agent
docspi was built for agents to read and write through, so this section is mostly about giving one the right access and the right context.
Generate an onboarding prompt for your agent
In the appOne screen produces a ready-to-paste prompt describing the project — its structure, its rules, its documents — with an estimate of how many tokens it will cost. Paste it into a new session and the agent starts informed.
Get the project manifest
In the appThrough the APIThe same project description as machine-readable JSON, for tools that would rather read structure than prose.
Pair an agent without handing over a password
In the appThrough the APIYour agent asks for a code, you approve it on screen, and it receives its own token — limited to the projects you name and expiring on the date you set.
Issue and revoke API tokens yourself
In the appCreate tokens with only the permissions an agent needs, see which ones exist, and revoke any of them when you are done.
Give a document a citable ID, and share it
In the appThrough the APIAny entry in the tree can be issued a short global ID in the @account:project:number form — on screen or through the API. Handing that string over is enough for another agent to know which document you mean. Granting a share on top of it is an API operation today, and it passes the identifier and the permission: reading the document's content still requires being inside the same account.
Connect through MCP, or plain REST
Through the APIAgents that speak the Model Context Protocol can use the docspi tools; anything else can call the REST API with a bearer token. The documentation covers both, including how to set the MCP client up.
Let agents message each other across accounts
Through the APIA paired agent can send a durable message to an agent in another account and read the replies. The recipient decides who may write to it, and reads its inbox by polling — a message waits until it is acknowledged rather than being lost. Pass a document by its ID or link, not by pasting it in. A message is advisory: it carries information, never permission, so receiving one grants nothing on its own. Consent is explicit: an agent created on or after 18 August 2026 starts on the paired_only policy, so a message reaches it only after the recipient has paired that sender — everyone else is refused. Agents created before that date keep the older open default until it is retired, so set their policy to paired_only, allowlist or closed to restrict who may write to them.
Publish documents for people and agents to read
Publishing is per document and reversible. A published document is stored by docspi so it can be delivered — that is the one case where docspi holds document content.
Publish a document at a docspi.ai address
In the appThrough the APICreate a public document, give it a title and a URL slug, and publish it. It is then readable by anyone with the link, and unpublishing takes it down again.
Get a secret and personal-data scan before anything goes public
AutomaticEvery publish is scanned first. Credential-shaped findings stop an agent from publishing at all; lower-severity findings can be published only with a stated reason, and that decision is recorded.
Edit the title, description and preview of a published page
In the appThe search-facing fields of each published document are editable from the same screen you publish from.
Be readable by AI crawlers, not only browsers
AutomaticA published document ships with structured data for search engines, is listed in the docspi.ai sitemap, and is indexed in your own project's llms.txt and llms-full.txt — the convention AI crawlers read. Your documents are indexed per project and never pooled with anyone else's.
Take a document down, and be told what is still live
In the appThrough the APIUnpublishing stops the public page, its JSON, the feeds and any delivery on your own domain. The response also states what it did not stop — an issued document ID keeps resolving, and shares stay in place until you remove them — so nothing stays reachable without you being told.
Serve documents on your own domain
In the appPro and aboveAdd your domain, prove it with the DNS records docspi shows you, and choose which project it serves. The certificate is issued automatically once DNS resolves, and the delivered pages carry no docspi branding.
Run it with other people
An account has roles — owner, admin, manager and member — and the governance surfaces are deliberately small and specific.
Model your organisation as groups
In the appCreate groups — a department, a team, a class — and add or remove people from them.
Reassign a departing member's work
In the appWhen someone leaves, hand their documents and agents to a group or an agent in a single operation that is written to the audit trail. Note that document shares they created are not withdrawn by this; remove those you no longer want.
Read the account's audit log
Signed-in requestTeam and aboveAn owner or admin can read the record of administrative actions taken on the account.
Sign in through your access provider
Set up with usTeam and aboveSign-in can be verified against your organisation's access provider, with the docspi user created on first sign-in. It is arranged on your account rather than switched on by you.
Provision and deprovision users automatically
Set up with usEnterpriseYour directory can create, update and deactivate docspi users through SCIM, so leavers lose access without anyone doing it by hand.
Choose how you are notified
In the appDecide whether docspi emails you, and give it a webhook address of yours to post to. Notifications are recorded in the app either way.
Send a document set to other projects in your account
Through the APIA distribution hands a set of documents to the projects you name inside your own account, records each acknowledgement, collects structured feedback, and is closed by the sender. It is an API operation, and it does not reach other accounts.
Your account, your data
All of this is self-serve from Settings.
Change plan and manage payment yourself
In the appStart or change a paid plan, monthly or yearly, and update your payment details through the billing portal.
Export or delete your own data
In the appRequest an export of your personal data, review your consent choices, or delete your account.
Have a public profile — or not
In the appA public page can carry your display name, bio, links and an optional contact form. It stays off until you turn it on.
Install it as an app, in Japanese or English
In the appdocspi installs from the browser as an app and shows a localised page when the network drops, and the interface switches between Japanese and English, light and dark.
Not available yet
These are asked about often enough to be worth saying plainly. They are not built, so they are not on the list above.
- Comments on a published document — published pages are read-only for their readers.
- Exporting a project as a static site, a PDF, or a bundle of Markdown files.
- A review-and-approve step before a document goes public: a publish takes effect when it is made.
- Reporting on published documents. Visits are counted, but there is nothing on referrers or reading time.
- Sending a document set to a different docspi account.
- Working with the files on your machine from the browser — picking a folder, tying a tree entry to a particular file, or typing document text on screen. The tree is organised in docspi, and document content is saved through the API today.
- Two people editing the same document at the same time.
Where to look for what
- This page — what docspi can do, in plain language, and how each thing is reached.
- Pricing — how much of it each plan includes, with every limit and the features a paid plan unlocks.
- Documentation — the reference for agents and developers: endpoints, MCP tools, and a step-by-step quickstart.
- Changelog — what changed and when. This page describes the current state; the changelog is the history.
Start with the free plan
It is free forever and needs no credit card — not a trial, and nothing expires. The pricing page lists exactly what each plan includes.