Pick tools by problem, not features
Tool lists usually fail because they read like feature inventories. You remember a tool when you remember the moment it saved you: the whiteboard you needed mid-interview, the JWT you pasted into a sketchy website at 1am, the API you had to poke at before the frontend existed.
So here are the five tools worth learning first, each mapped to the problem it solves, then seven more that quietly earn their place in a daily workflow. Everything below is free to use, and most of it is open source.
1. Excalidraw — architecture and visual thinking
Excalidraw is an infinite whiteboard for sketch-style diagrams: architecture, wireframes, flows, entity relationships, brainstorming. The value is not that you can draw boxes — every tool draws boxes. It is that the deliberately rough visual style signals this is a thinking artifact, not a specification, so nobody argues about alignment while you are still arguing about the design.
The two-minute version of a diagram you would otherwise spend twenty minutes on:
Where it earns its place
- System design interviews. Shared canvas, no setup, exports cleanly if you want to keep the artifact.
- Explaining bugs. A five-box sequence beats four paragraphs in a thread.
- API and schema design. Draw the entities and relationships before writing the migration.
- Documentation. Export to PNG or SVG, keep the editable
.excalidrawfile next to it so the diagram stays maintainable.
.excalidraw source alongside the exported image in your repo. An exported PNG nobody can edit becomes wrong documentation within a quarter.2. DevToys — the developer Swiss Army knife
DevToys replaces the reflex loop of searching for "json formatter online", "base64 decoder", "jwt debugger" and pasting work data into whichever ad-covered page ranks first that week. It bundles roughly thirty utilities into one desktop app that runs locally.
Google → JSON formatter → paste production payload Google → Base64 decoder → paste internal token Google → JWT decoder → paste customer's auth token ← this one is a real problem Google → UUID generator Google → Hash generator
The tools you will actually open
- JSON / YAML formatter and converter — turn a single-line log dump into something readable, or move config between formats.
- JWT decoder — inspect the header and payload of a token to see claims, issuer and expiry.
- Base64 encode / decode — for data URIs, basic auth headers and encoded payloads in logs.
- Hash and checksum generator — MD5, SHA-1, SHA-256 fingerprints to compare files or verify a download.
- UUID generator — stop hand-typing test identifiers.
- Regex and JSONPath testers — iterate on a pattern with live match highlighting instead of redeploying to find out.
{"user":{"id":123,"name":"John","skills":["Java","Redis"]}}
{
"user": {
"id": 123,
"name": "John",
"skills": ["Java", "Redis"]
}
}3. Hoppscotch — API testing without the account wall
For backend work this is the highest-value tool on the list. Hoppscotch sends HTTP requests and shows you the response — method, headers, query parameters, body, auth — with a fast UI and no mandatory sign-in before you can send your first request.
POST /users HTTP/1.1
Host: api.example.com
Authorization: Bearer <token>
Content-Type: application/json
{
"name": "Upasana",
"role": "engineer"
}The features that matter past the first request
- Collections. Group related requests so the API surface is documented by usage rather than by a stale wiki page.
- Environments. Define
API_URLonce per environment and switch between local, staging and production instead of editing every URL. - Authentication. Basic auth, bearer tokens, OAuth 2.0 and API keys are supported directly.
- Pre-request scripts. Generate a timestamp, sign a payload or fetch a token before the request fires.
- Tests. Assert on status codes and response fields so a request becomes a repeatable check.
- CLI. Run saved collections from a terminal or a CI pipeline — the step that turns manual poking into a regression suite.
User APIs
├── GET {{API_URL}}/users
├── GET {{API_URL}}/users/:id
├── POST {{API_URL}}/users
├── PUT {{API_URL}}/users/:id
└── DELETE {{API_URL}}/users/:id
Local API_URL = http://localhost:8080
Staging API_URL = https://staging.example.com
Production API_URL = https://api.example.com4. DevDocs — documentation without twenty tabs
DevDocs pulls official documentation for many languages, frameworks and platforms into one fast search interface. The point is not to avoid documentation — it is to reach the authoritative page in one keystroke instead of routing through a search engine, a content-farm blog and a nine-year-old Stack Overflow answer.
The underrated feature is offline mode: DevDocs can install documentation sets into browser storage using IndexedDB and a service worker, so your reference material survives a dead connection on a train or a locked-down network.
How to get value on day one
- Enable only the doc sets you use, so search results stay relevant.
- Learn the keyboard shortcuts — the tool is built around typing, not clicking.
- Install offline docs for your primary language before you travel.
- Add it as a browser search keyword so a lookup starts in the address bar.
5. ExifTool — the metadata your files carry (bonus)
This one is more niche than the first four, which is why it belongs as a bonus rather than a core pick. But it produces the most "I had no idea" reaction: files carry far more information than the bytes you can see.
$ exiftool photo.jpg File Name : photo.jpg File Type : JPEG Image Width : 4032 Image Height : 3024 Make : Apple Model : iPhone 14 Pro Create Date : 2026:03:11 18:42:07 Software : 17.4.1 GPS Latitude : 12 deg 58' 21.6" N ← only when the file actually stores it GPS Longitude : 77 deg 35' 41.2" E
GPS coordinates only appear when the capturing device recorded them and nothing stripped them since. That conditional is exactly why the tool matters for anyone shipping user-facing uploads.
When you will actually need it
- Photo upload pipelines. Strip location and device metadata before storing or serving user images — this is a privacy requirement, not a nice-to-have.
- Media processing. Read orientation tags so images do not render sideways after a resize.
- Digital asset management. Extract capture dates and camera data for cataloguing.
- Incident and forensics work. Establish provenance of a file.
# remove every metadata tag before storing a user upload exiftool -all= -overwrite_original upload.jpg
7 more free tools worth installing
The five above cover design, utilities, APIs, docs and files. These seven fill the gaps you hit in a normal week.
Where DevToys gives you one utility at a time, CyberChef lets you build a pipeline: URL-decode, then Base64-decode, then gunzip, then pretty-print JSON — all in one recipe you can save and share. It is the right tool for a mangled payload of unknown shape. Run the offline build for sensitive data.
Once JSON is in your terminal, jq filters, reshapes and aggregates it without a throwaway script. It composes with curl and log tooling, which is what makes it stick.
curl -s api.example.com/runs \
| jq '.runs[] | select(.status == "failed") | {id, error, duration_ms}'The terminal counterpart to Hoppscotch. JSON bodies, sane defaults and colourised output without remembering six curl flags.
http POST api.example.com/users name=Upasana role=engineer
Its real feature is the right-hand pane that explains each token of your pattern, plus flavour switching between PCRE, JavaScript, Python and Go. Writing regex against a live match set with an explanation beside it turns a guessing game into a readable exercise.
Type */15 9-17 * * 1-5 and read back "at every 15th minute past every hour from 9 through 17, Monday through Friday". Thirty seconds here prevents a job that silently runs once a month instead of every fifteen minutes.
For merge conflicts and config drift, a side-by-side visual diff beats reading conflict markers in a terminal. It handles whole-directory comparison too, which is the fastest way to answer "what is different between these two environments?".
A single client across every database you touch, with schema browsing, an ER diagram view and query history. The ER view alone is worth it when you inherit a schema nobody documented.
Which tool for which problem
| Tool | Solves | Runs | Reach for it when |
|---|---|---|---|
| Excalidraw | Diagrams & architecture | Browser / desktop | You are explaining a design to a human |
| DevToys | Everyday conversions | Local desktop | You would otherwise paste data into a random website |
| Hoppscotch | API development | Browser / desktop / CLI | The endpoint exists but the UI does not |
| DevDocs | Documentation lookup | Browser / offline | You need the official answer, fast |
| ExifTool | File metadata | CLI | You handle user uploads or media files |
| CyberChef | Multi-step decoding | Browser / offline build | The payload is encoded three ways |
| jq | JSON in the terminal | CLI | The response is JSON and you want three fields |
| HTTPie | Quick HTTP calls | CLI | You want curl without the flags |
| regex101 | Pattern building | Browser | The regex almost works |
| Crontab.guru | Schedule syntax | Browser | Before you ship any cron expression |
| Meld | Diff & merge | Desktop | Conflict markers stopped making sense |
| DBeaver | SQL across engines | Desktop | You touch more than one database |
Building your own free toolkit
Do not install all twelve today. Adoption fails when a tool has no attached trigger. Pick the problem that annoyed you most this week and install only that tool, then let the next annoyance choose the next one.
- Prefer local over web for anything sensitive. Tokens, customer payloads and internal identifiers should not travel to a formatter you found through search.
- Prefer tools with a CLI. Anything scriptable can eventually run in CI, which is how a manual habit becomes an automated guarantee.
- Keep artifacts in the repo. Diagram sources, API collections and query snippets belong next to the code they describe.
- Learn one keyboard shortcut per tool. That single shortcut is usually the difference between a tool you installed and a tool you use.
Tooling will not compensate for weak fundamentals — no diagram tool teaches you where to place a cache, and no API client tells you which consistency guarantee your write path needs. What good tools buy you is the removal of friction, so the time you spend goes into the design decision rather than the mechanics around it.