The API client your team has been waiting for
An open-source desktop client for HTTP, gRPC, GraphQL, and WebSocket. A tool, not a platform.
30 MB · WebView · MIT
- 30 MBInstall size
- <500 msCold start
- 0Telemetry endpoints
- 4Protocols
HTTP · gRPC · GraphQL · WS
Request.
Staging is up, standup is in five minutes, and you just need one clean answer. Open the client, send the request, watch the green 200 appear, check the status bar, skim the raw bytes. That's the whole flow. While heavy clients are still loading, you're already done with the first request and on to standup.
- Median round-trip, staging
- 41 ms
Schema reflection
Schema.
The useful part of gRPC starts when you stop hunting through proto files. Drop in the endpoint, let reflection pull the schema, then click into unary, server-stream, client-stream, or bidi. One click gives you an example payload with the right shape, so you can test the method instead of reconstructing it.
- Proto to populated body
- 1 click
Sandboxed JS
Scripts.
This is for the boring useful stuff: stamp a token before send, check a field after response, fail loudly when a contract drifts. Scripts run in a locked-down Goja sandbox. No files to poke at, no network to phone home, five seconds max. The result shows up right beside the response.
- No network, no files
- Sandbox
Self-hosted sync
Sync.
We treat your API work as your data, not ours. That's why sync is opt-in and self-hosting is a first-class path. Put the sync server on your VPS, secure it with your TLS setup, keep SQLite as the source of truth, or leave sync off entirely. Your team decides where data lives and how it moves.
- Your server, your rules
- Opt-in
Beyond the basics
And that's not all
MCP server
Agents and LLM tools call your collections over MCP.
Environments & variables
Switch dev, stage, and prod without editing requests.
Import from Postman
Collections and environments move over in a couple of clicks.
Cookies
The cookie jar keeps sessions between requests automatically.
Workspaces
Keep projects and teams apart without mixing collections.
Keyboard-first
Send, switch tabs, find a request — without touching the mouse.
Demo
Try it without installing
The client opens right in your browser — no signup, no install, no email. Same interface, sample data.
Community
Built in the open
No invented download counters, no logo walls for gravitas. Here is what actually exists today — and it grows with every release. Pre-v1.0 is the best time to influence the product.
- MITCode license
- 40Client releases in 2026
How it compares
| Tetiva | Postman | Insomnia | Bruno | Yaak | |
|---|---|---|---|---|---|
| Install size | ~30 MB | ~140–160 MB | ~350–580 MB | ~100–160 MB | ~40–77 MB |
| Account | Not required | Encouraged, lightweight mode optional | Scratch Pad only without an account | Not required | Not required |
| Telemetry default | None | Yes; in 2025 sent full URLs despite privacy settings | Yes (opt-out) | None | None |
| Built-in sync | Self-hosted realtime (optional) | Cloud only | Cloud + Git, server is closed | Git/files | Git/files |
| Protocols | HTTP / gRPC / GraphQL / WS | HTTP / gRPC / GraphQL / WS | HTTP / gRPC / GraphQL / WS / SSE | HTTP / GraphQL / gRPC / WS | HTTP / gRPC / GraphQL / WS / SSE |
| MCP server for AI agents | Built into the client | No | MCP client (connects to external servers) | No | Plugins |
| Scripts sandbox | Sandboxed JS (no FS, no network, 5s) | Postman JS | Insomnia JS | Inline JS | Plugin API |
| License | Client — MIT, sync server commercial | Commercial | Apache 2.0 (client), sync server closed | MIT (core), Golden Edition commercial | MIT + paid binary licenses |
| Built with | Wails + Go + Vue | Electron | Electron | Electron | Tauri + Rust + React |
Sizes from official distributions as of 2026-06-10 (vary by OS). Verify before relying on for tool decisions.
Pricing
What will cost money
Free
forever$0
- All 4 protocols
- Local workspaces
- Scripts and environments
- Import and export
Pro
for teamsat release
- Cloud sync
- Organizations and shared workspaces
- Roles and permissions (RBAC)
Self-hosted Community
free, in your perimeter$0
- Your own sync server
- Data never leaves your network
- Up to 5 users
- Basic roles, community support
Self-hosted Business
for companiesat release
- No user limit
- SSO (SAML / OIDC)
- Audit log and fine-grained RBAC
- Priority support
Prices land with the release. The beta is free for everyone.
Frequently asked questions
- Is Tetiva free?
- Yes. The desktop client is free and open-source under the MIT license. The optional sync server will have a commercial license once v1.0 ships.
- Do I need an account to use it?
- No. The desktop app opens to an empty request and lets you send HTTP, gRPC or GraphQL calls without ever creating an account or signing in. An account is only required if you choose to connect to a sync server.
- Does it phone home?
- No. The client only makes the network requests you ask it to. No analytics, no telemetry, no usage reporting are sent by default. The source is on GitHub if you want to verify.
- Can I import my Postman collections?
- Yes. Postman v2.1 JSON exports import directly. Folders, requests, headers, body types, auth, file fields and scripts are preserved.
- Is the sync server required?
- No. The sync server is fully optional. You can run Tetiva as a single-user desktop app indefinitely. When you do want sync, you can self-host the server with a Docker Compose template.
- How does this compare to Bruno or Yaak?
- Bruno and Yaak store collections as files and sync them through Git — great for version control, but there is no live sync: a teammate's changes arrive with the next pull. Tetiva takes a different path: SQLite locally + an optional self-hosted sync server with realtime updates, plus a built-in MCP server that lets AI agents manage your collections.
- Is there a CLI?
- A standalone MCP server CLI ships with the desktop app for AI agents and tests. A general-purpose CLI for CI/CD pipelines (similar to Newman) is on the roadmap but not in MVP.
- How hard is self-hosting the sync server?
- A Docker Compose template is provided. The server requires MongoDB as a Replica Set (single-node mode supported via init script). For small teams, total setup is around 20–30 minutes on a fresh VPS.