Now Available — v1.8.0

Teaspoon IDE

Designed to get real work done on a teaspoon of tokens.
A standalone, privacy-first, Electron-based AI coding agent IDE that bundles everything you need to build software.

Download View on GitHub

Download — v1.8.0 (2026-11-21)

The links below download files directly from GitHub Releases (no GitHub account needed). Older versions are on the Releases page.

FilePlatformSHA256
teaspoon-ide-1.8.0-win-x64.zip Windows x64 (portable: unzip and run Teaspoon.exe) 5f1e1d61341bb943044a60224a051a76c074f3718ae20bac64a3d824dc682576
teaspoon-ide_1.8.0_amd64.deb Linux x64 (Debian / Ubuntu) 2d8450fb4b52bd38e220dee3f8e36b1d423a09491011aa30ee6658a7778953c3

Warning at launch and how to verify

Teaspoon IDE is not code-signed yet, so Windows SmartScreen may show a warning ("Windows protected your PC") when you run it. You can verify the file is identical to the distributed one by comparing its SHA256 hash with the table above.

# Windows (Command Prompt / PowerShell)
certutil -hashfile teaspoon-ide-1.8.0-win-x64.zip SHA256

# Linux
sha256sum teaspoon-ide_1.8.0_amd64.deb

If the printed hash matches the table, the file has not been tampered with.

Highlights

  • All-in-one File explorer, Monaco editor, AI chat, Git, and a true TTY terminal in a single app. No more jumping between tools — development happens entirely inside Teaspoon IDE.
  • Autonomous agent loop The AI lists, reads, searches, and edits files across multiple steps, feeding results back until the task is done — then summarizes what it did in plain language. With Ollama, replies stream token by token, and each message shows the model used and its response time.
  • Safe write control Writes are confined to the project root. File creation, diff edits, and command execution all go through approval dialogs, and checkpoint/rollback lets you undo only the AI's changes.
  • Context optimization Sends the file tree first; the AI fetches file contents on demand via GREP / READ_FILE. Dramatically reduces token usage per message.
  • Keys stay yours Works through a LiteLLM proxy so real API keys never live on the client. Ollama support enables fully offline operation. File paths sent to the AI are project-relative, so absolute paths containing your OS user name never leak to the cloud.
  • A free BYOK endpoint Teaspoon IDE itself is free. It follows a BYOK (Bring Your Own Key) model — you supply your own Gemini API key, so there is no subscription fee and no markup on API usage. You simply pay Google directly for what you use.
  • Low-cost — or completely free Focused on Gemini's inexpensive Flash models, so pay-as-you-go costs stay minimal (Gemini 3.8 Flash is offered at an introductory price, half the standard rate, through the end of 2026). Or choose Ollama with a local LLM like Gemma 4 for a completely free, offline, secure AI coding agent IDE that never sends your code off your machine.
  • Fine-grained local-LLM control (v1.2.0+) Ollama-only settings for CPU Threads (a cap on inference threads), Thinking (reasoning on/off plus Low/Medium/High effort levels), Context Size (how much history the model keeps), and Temperature (output variance) live right in Settings. Keep your PC responsive during generation, and skip the thinking trace for much faster replies — controls only possible on Ollama's native API, which OpenAI-compatible tools like Cline don't expose. Read the measured-effect article.
  • Practical agent loops on small local models (v1.8.0+) Recovery logic keeps small models' grammar slips from ending a run — edits that were already applied no longer false-alarm, failed replies stay out of the history sent back to the model, and truncated commands are detected and re-requested. In real-machine testing, Ollama with gemma4:e4b (an edge-class model, ~4.5B effective parameters) completed a four-file game generation in a single agent run. Slower than cloud models, but verified to reach the goal — fully offline and free.
  • Ready for teams (v0.5.0+) Flip on "Require organization sign-in" for centrally managed use: members just sign in with an issued account — no API keys to hand out. Remaining budget shows as a percentage, and usable models are governed by the organization. See Teaspoon IDE Organization.
  • Production-grade polish Git covers stage/commit/push/pull plus GUI dialogs for cloning, remotes, initial commits, upstream-tracked pushes, and user config. Diff view, project-wide search, Markdown preview and PDF export, and fourteen themes including Organic Light are built in. Ships as a portable exe and a .deb package.

Tech Stack

Electron / React / TypeScript / Monaco Editor / Gemini API & Ollama / Vite

License: FSL-1.1-MIT (source-available — see "About the License" below)

For real usage examples, see Teaspoon IDE Example (Othello) and Teaspoon IDE Example 2 (falling-block puzzle — playable in your browser). For setup instructions (Gemini API key, Ollama local LLMs, remote Ollama connections, etc.), see the Teaspoon IDE Settings Guide. For centrally managed deployments (budget caps, model governance), see Teaspoon IDE Organization.

See the CHANGELOG for what's new in each release.

Feedback and bug reports are welcome via GitHub Issues.

About the License (FSL-1.1-MIT)

Teaspoon IDE is distributed under FSL-1.1-MIT (Functional Source License), a "source-available" license. It differs slightly from the MIT open-source license, but ordinary use is free and unrestricted.

You are free to

  • Download and use it — for personal and business use alike
  • Modify and customize the source code to your liking
  • Deploy it as an internal tool within your company
  • Fork and redistribute it, as long as the license terms and copyright notices are kept

You may not

  • Use it as a commercial product or service that competes with Teaspoon IDE — e.g. selling it as-is, selling a renamed clone, or offering it as a paid service
  • Remove the license terms or copyright notices when redistributing
  • Use the "Teaspoon IDE" name or trademarks for any purpose other than identifying its origin

Each release automatically converts to the MIT license two years after publication, after which even commercial use is fully permitted. See the LICENSE file in the repository for details.