PaperBoy

PaperBoy is a Rust-native alternative to Postman. Collections and environments are plain, git-friendly text files — Hurl .hurl files for requests, .vars files for environments — so everything can be committed, diffed, and shared through a normal git repository. There is no hosted service and nothing runs outside your machine.

It ships as a single binary with two front-ends:

• Terminal UI (default) — a full interactive client for building, editing, running and organizing requests.
• Headless CLI (-c/--collection) — runs a collection end-to-end and prints the results, for scripts and CI.

Contents

• Installing & running
• Core concepts
• Features
• Working with collections & environments
• Loading & saving via git
• Secrets & environment variables
• Keyboard shortcuts
• Headless CLI mode

Installing & running

From the paperboy directory:

cargo run            # launch the terminal UI
cargo build --release

Core concepts

• Collection — a .hurl file: an ordered list of requests, each with a method, URL, headers, cookies, body/form fields, and optional [Captures]/[Asserts] sections. Postman collection exports (.json) can also be opened directly — they are imported and converted to Hurl automatically.
• Environment — a .vars file: KEY=value lines supplying the {{ VAR }} values a collection references. A value can be a literal, or a reference to a secret provider (see Secrets).
• Scratch Space — the first, always-present tab. It behaves like any other collection (add/run/edit requests) but isn't tied to a file until you explicitly save it — a place to try things out without committing to a collection first.

Features

• Multiple collection tabs, each with its own environment.
• Add, edit, reorder, run, and delete requests within a collection.
• Edit Request wizard: pressing Enter on a request opens the same form used for "New Request", prefilled with its current method, URL, headers, cookies, body, form fields, [Captures], and [Asserts] — no manual JSON or Hurl syntax required. Shift+R opens Raw Mode instead, a direct editor for the request's real Hurl text (for anything the form doesn't expose, such as query params, Basic Auth, or the expected status); saving reparses the text and applies it back to the request, keeping invalid text open for correction rather than discarding it.
• [Form]/[Multipart] and [Cookies] sections in the request wizard: add form fields (enabled checkbox, Key, a Kind dropdown defaulting to Text — press Enter to open it and pick File instead, Value, a separate Content-Type column, and a free-text Description column — Kind comes before Value so filling a row naturally picks Text/File before typing the value) and cookies (checkbox, Key, Value), using the same table and tabbing conventions as [Headers]. Saving picks the correct Hurl section automatically — all-Text fields become [Form], and any File field promotes the section to [Multipart]. File fields are colored to show whether the referenced path exists and is readable — relative paths resolve against the collection's own directory, matching where Hurl actually looks for them when sending the request. With focus on a File-kind Value cell, Ctrl+F (or Enter) opens a file picker to choose the path instead of typing it (without stealing focus to the Content-Type column afterwards); the Content-Type cell is a dropdown offering "Auto" plus the fixed MIME types from hurl.dev (still editable as free text for anything else), and is auto-set from the picked file's extension. An empty/auto-detected Content-Type cell shows a dimmed "Auto" placeholder rather than looking blank. Editing a request into an impossible Body + Form/Multipart combination shows a clear status-bar error instead of sending it.
• Headers, Cookies, Form, Asserts, and Captures all start empty — just a "+ Add …" button for each, with no default blank row to delete for requests that don't need one (a lingering blank row used to intercept the ↓ key meant for scrolling past that section). To remove a row you no longer need, focus any cell in it and press Ctrl+D; the wizard's hint bar shows ^D delete row as a reminder whenever focus is on a deletable row.
• Enabled checkbox is reachable with ← from a row's Key cell — it's the leftmost visual column of every Headers/Cookies/Form row, so arrow navigation now matches that layout instead of requiring a trip through every other column to reach it. A brand new row still starts focus on Key (not the checkbox), matching how you naturally fill a row in. Ctrl+E toggles the focused row's enabled state and jumps focus to the checkbox, from anywhere in that row.
• Dropdowns only auto-open when their cell is empty. The Key suggestion and Content-Type dropdowns pop open automatically the first time you land on an empty cell, but once a cell has a value, navigating onto it again leaves the dropdown closed — press Enter to bring it back up without disturbing the current value. The Form Kind dropdown follows the same "populated cell" rule: since it defaults to Text on a fresh row, it never auto-opens — Enter always reveals it to switch to File. This keeps ↑/↓ navigation through a populated section from getting stuck reopening a dropdown on every row.
• Scrollable, navigable tables: when a Headers/Cookies/Form/Asserts/ Captures section has more rows than fit on screen, it grows to show up to 5 rows before a scrollbar appears on the left edge, right next to the data; ↑/↓ scroll the focused row into view, and Ctrl+↑/Ctrl+↓ jump straight to the first row of the previous/next section. The "+ Add …" row always stays reachable even when the list is scrolled, and a long Body also grows a left-hand scrollbar once it overflows.
• Per-section wizard tabs: the request wizard has a tab bar (All │ Headers │ Cookies │ Form │ Body │ Asserts │ Captures), switched with PageUp/PageDown and reordered with Ctrl+Shift+←/→ (just like collection tabs). Picking a section tab gives that section almost the whole dialog to itself — handy for editing long Headers or Form-field lists — while All keeps the combined, everything-at-once layout. Every tab is just a different view of the same request, so edits made on one tab show up immediately on every other. Switching to a section tab jumps focus straight to its first entry (Body's tab drops you right into editing), and Tab/Shift+Tab/↑/↓/Enter are confined to the active section so navigation never jumps to a section you can't see.
• Request chaining: capture a value from a response (Hurl [Captures]) and reference it as {{ name }} in any later request in the same collection.
• Basic Auth support (Hurl's native [BasicAuth] section).
• Colored, substituted request preview: {{ VAR }} placeholders in the Request JSON panel are shown replaced with their real value, colored by status (green = loaded, cyan = literal, orange = still loading, red = missing) — while the editor itself always keeps the original {{ VAR }} text, so you always know exactly what will be sent versus what's stored. Secrets substituted into the preview are masked as 8 dots.
• Editing an environment entry to look like a secret reference (e.g. typing {{ op://vault/item/field }} or {{ ssm:/path }} into a value) triggers an automatic load attempt, the same as if it had been in the file originally.
• Load/save collections and environments from local files, or straight from a git remote (see below) — no local clone required. A collection or environment loaded from git shows a small ⎇ icon in its tab title / Environment heading, and can be pushed back to a new branch or tag with Save Collection to Git….
• A recently-used git URLs dropdown in the "Load from Git" wizard, most recent first.
• Resizable panes: the left column width and the response pane height are both adjustable and persisted across restarts.
• Tab management: close (Ctrl+W/x), reopen the last-closed tab (u), cycle tabs ([/] or PageUp/PageDown), and reorder tabs (Ctrl+Shift+←/→).
• Deleted requests can be restored too. Pressing x on the Requests list deletes the selected request; u on that same pane brings back the most recently deleted one (last-deleted-first, one collection's stack per collection) — the exact parallel of reopening a closed tab, just scoped to individual requests instead of whole collections.
• The app remembers your window layout, active tab/request, language, and recently-used git URLs across restarts.
• English, French and Danish UI languages (Options → Language).
• Nested folders. A request's name can encode a folder path with / separators (e.g. Auth/Tokens/Refresh); the Requests list browses these as folders — Enter on a folder row descends into it, Enter on the "‹ .." row (or Backspace) goes back up — instead of an ever-more-indented tree, and the panel title shows a breadcrumb of the folder you're in. Importing a Postman collection preserves its nested folder structure this way automatically; native Hurl collections get the same behavior for free simply by naming requests with / in them. Creating a new request while browsing a folder prefills the Name field with that folder's path.

Working with collections & environments

Open the File menu with f. Load and save actions are grouped together:

Notes on saving:

• Save Collection / Save Environment write back to the file the tab was originally loaded from. If there are no new or modified entries, saving is a no-op and no confirmation is shown (there's nothing to warn about). If there are unsaved changes, you'll get a confirmation naming how many new/modified requests (for a collection) or variables (for an environment) will be written — the two counts are tracked and reported independently, so saving a collection only ever mentions collection changes, and saving an environment only ever mentions environment changes.
• Save As… always prompts for a filename. If the chosen path already exists you'll be asked to confirm the overwrite (Yes/No) before anything is written.
• The Scratch Space (tab 0) can be saved like any other collection — pick Save Collection As… to give it a file for the first time.
• A modified request shows a pencil icon next to it in the Requests list until it's saved.

Loading & saving via git

This is the one workflow that isn't just "open a file", so here's the full step-by-step. Loading fetches a single file out of a remote repository without cloning it: it lists the repo's branches/tags, fetches just enough git history to read the file tree for the ref you pick, and checks out only the one file you select. No other file in the repo ever touches your disk.

1. Press f to open the File menu, then choose Load from Git… (for a collection) or Load Environment from Git….
2. Connect — type the repository URL (https://… or git@…). If the repo needs an access token, Tab to the token field and enter it (only used for this fetch — the wizard supports GitHub-style https://x-access-token:<token>@host/... URLs and injects the token for you). Press ↓ on the URL field to open a dropdown of your recently-used URLs and pick one instead of retyping it; press Enter to connect.
3. Pick a ref — choose a branch or tag from the list (type to filter).
4. Pick a file — choose the .hurl/.json (or .vars) file from the ref's file tree (type to filter).
5. The file is checked out and opened as a new tab. Its URL is remembered at the top of the "recently used" list for next time, and a small ⎇ icon is shown in the tab title (or, for an environment, in the Environment panel heading) to mark it as loaded from git.
6. When loading a collection, you're then asked whether to also load its environment — answering Yes reuses the same file listing already fetched in step 4 (no second network round-trip) and lets you pick the .vars file to pair with it. Answering No (or loading an environment on its own via Load Environment from Git…) leaves it as-is.

Loading a directory of files, or importing a whole repo, is not supported this way — for that, see Loading a Workspace from git below.

Loading a Workspace from git

A Workspace is a folder containing many .hurl/.json collection files, browsed and chosen from through a single tab (press w on a Workspace tab to reopen its file-tree picker at any time). Workspaces can also be loaded straight from a remote repository, instead of only from a local folder:

1. Press f to open the File menu, then choose Work(s)pace from Git….
2. Connect and pick a ref, exactly as for a single collection/ environment above.
3. Choose which files to download — a repo can hold plenty of large, unrelated files that have nothing to do with your collections, so before anything is fetched you pick one of: .hurl and .json files only (the default), .hurl files only, .json files only, or every file. Only files matching this choice are ever downloaded — anything else in the repo never touches your disk.
4. The matching files are checked out into a throwaway local folder. Before the new tab is created you're asked whether to keep it temporarily (the folder above) or save it to a permanent location now — picking the latter opens a folder browser and a name prompt, then copies the files there immediately and binds the new tab to that permanent folder instead. Either way, the file-tree picker then opens immediately so you can choose which collection to view.

⚠️ A temporarily-kept Workspace's files are not cleaned up automatically. Unlike the single-file collection/environment load above, if you choose to keep it temporary its files stay on disk in a temp folder for as long as its tab exists — including across closing and reopening the tab within the same session (u / Ctrl+Shift+T undoes a close, so the folder can't be safely deleted the moment the tab closes). Restarting PaperBoy does not delete it either. If you load the same Workspace from git repeatedly, or load many different ones and keep them all temporary, these folders will accumulate under your system's temp directory and you may want to clear them out yourself from time to time — or avoid the problem entirely by choosing save it to a permanent location when asked, or later via File → Save → (W)orkspace… (available for any Workspace tab, local or from git; it copies the whole folder to a destination and name you choose, and stops tracking it as a temporary git download). Saving a Workspace-loaded collection back to git is also not currently supported — use Save Collection As… into your own local clone instead.

Saving to git

A collection that was itself loaded from git (shown by the ⎇ icon in its tab title) gains a Save Collection to Git… File menu item, which pushes a new commit straight to the remote — no local clone, staging area, or manual git commands needed. Like loading, this never performs a full checkout: only the file(s) you're actually writing are ever fetched or touched, no matter how large the repository is.

1. Connect — the repository URL is pre-filled with the one this collection was loaded from (editable, e.g. to push to a fork), plus an optional access token field.
2. Choose what to save — the collection's in-repo path (defaults to where it was loaded from), and, if an environment is attached, whether to also save it in the same commit (and its in-repo path).
3. Choose a target — Tab switches between Branch and Tag:
   • A branch name defaults to the one you loaded from (so pressing Enter here just appends a new commit to it); typing a different name targets another existing branch (also a plain append) or creates a brand-new one. ↓ opens a dropdown of the remote's existing branches to pick from instead of typing. No merge or rebase is ever attempted — if the branch has moved in a way that makes the push a non-fast-forward, it's simply reported as an error.
   • A tag name must be brand new. PaperBoy re-fetches the remote immediately before checking, so even a tag another user created seconds ago is caught — an existing tag name is always rejected and never overwritten, with no way to force it.
4. Commit message — auto-generated (Update <name> via PaperBoy) and editable before pushing. The commit author is your system git identity (git config user.name/user.email), or a generic PaperBoy <paperboy@localhost> identity if that isn't configured.
5. Once pushed, a branch-target save updates the collection's (and, if included, the environment's) remembered git origin to that branch and clears their "new"/"modified" markers, exactly like a local Save. A tag-target save clears the markers too but leaves the remembered branch origin untouched, so the next "Save to Git" still defaults back to your working branch.

For a collection that wasn't loaded from git (or when you'd rather manage git yourself), the old workflow still works exactly as before: use Save Collection As… / Save Environment As… and choose a path inside your own local clone of the repository, then commit and push with your normal git tooling outside the app.

Secrets & environment variables

An environment .vars file has one KEY=value entry per line. The value can be:

Provider-backed values (op://…, ssm:…) are resolved in the background when the environment loads; the Environment panel shows their status while this happens. Every 1Password reference across every open collection's environment is resolved with a single op inject call (one authorization prompt), instead of one prompt per collection. If an entry fails to load (missing tool, declined auth, unreachable parameter store), select it in the Environment panel and press r to retry just that one entry — works for process env vars, 1Password, and SSM references alike. When editing such an entry, a "still secret?" checkbox lets you keep it masked or, if you're confident the new value isn't sensitive, save it as a plain value in the state. Rewriting a value to look like a provider reference (typing {{ op://... }} or {{ ssm:... }} into any entry) automatically triggers a load attempt for it, same as on initial file load.

Keyboard shortcuts

Open the in-app help overlay any time with ? or F1 for the full, up-to-date list. Highlights:

Headless CLI mode

Run a collection end-to-end from the command line, with no UI:

paperboy -c collection.hurl
paperboy -c collection.hurl -e environment.vars

• -c/--collection accepts a Hurl .hurl file or a Postman collection export (.json), imported automatically.
• -e/--env supplies a .vars environment file for any {{ VAR }} references in the collection (required if the collection uses {{ BASE_URL }} or similar).
• -b/--batch runs every request as a single Hurl call instead of the default streaming mode (see below).

By default, each request's method, URL, status, asserts, captures, and body (truncated) are printed as soon as that request finishes, instead of waiting for the whole collection — output is color-formatted (skipped automatically when not writing to a terminal, or when NO_COLOR is set) to make passes, failures, and sections easier to tell apart at a glance. Streaming reuses the same hurl crate runner one request at a time, so captures still chain correctly between requests — but it can't carry Hurl's automatic cookie jar (cookies remembered from Set-Cookie response headers) across requests the way running the whole collection in one call can; a warning to this effect is printed at startup. An explicit [Cookies] section on a request is unaffected either way. If a collection relies on cookie continuity between requests, pass --batch to run it as a single call like before, trading incremental output for that guarantee.

The process exits 0 if every request passed, non-zero otherwise.