RRAME Code
VS Code extension · Apache-2.0

RAME Code

A local-first coding agent for VS Code, powered by Meta Muse models. Every action needs explicit permission, and the record of what happened is written by the host, not by the model.

Download v0.1.0 (.vsix) GitHub

Public preview. Try it on a disposable project: it is not a production security boundary.

Published in v0.1.0

Only what the release does today.

What works today

  • Agent, Mission, Chat

    Three modes in a single Activity Bar view.

  • Muse Glimmer, local

    Runs through a loopback-only OpenAI-compatible llama-server endpoint.

  • Muse Spark and Muse Wonder

    Through Meta Model API, with your own key stored in VS Code SecretStorage.

  • Bounded tools

    Workspace inspection, file operations, patches and allowlisted test commands. No arbitrary shell tool.

  • One-use approvals

    Each approval is bound by digest to the exact action, thread, provider, workspace and file preconditions.

  • Path and secret protection

    Blocks absolute paths, traversal, symlink escapes and selected secret and control paths.

  • Host-written action record

    Kept separate from what the model says, so the record is evidence and not a claim.

  • Persistent threads

    Local threads, cancellation, per-provider usage accounting and explicit consent before anything goes from local to cloud.

  • Generation controls

    Reasoning effort, context and output budgets, validated per provider and model. Cut-off responses are discarded.

AskEvery tool action waits for approval. Default.
Safe autoPolicy-checked reads run on their own; writes and tests still ask.
Workspace fullAllowed workspace actions run without prompting. Higher risk.

Install

Download the VSIX, then in VS Code open Extensions, the … menu, Install from VSIX. Or from a terminal:

code --install-extension rame-code-0.1.0.vsix

Known limits

  • Not a production security boundary. Use disposable projects.
  • Project test scripts run with the operating-system permissions of VS Code.
  • The local audit is bounded but not tamper-proof.
  • No persistent project memory and no mission resume after a full restart yet.
Roadmap

Nothing here is part of a published release.

What comes next

PlannedIn testing · not releasedMoves to Published only after a verified release.
  • Planned

    Versioned project decisions

    User-authorized decisions with history, whose dependent work can be flagged as stale when a decision changes.

  • Planned

    Stale-artifact detection

    Make visible which files and plans were built on a decision that no longer holds.

  • Planned

    Missions that survive a restart

    Resume a mission after VS Code or the extension restarts completely.

  • In testing · not released

    Local-to-cloud cascade

    Glimmer works locally, asks Meta for advice only with consent, then Glimmer verifies locally. Under test, in no release.

  • In testing · not released

    Self-improvement runs

    RAME Code using its own Agent mode to inspect and improve RAME Code. A release will say so explicitly once it is verified.

v0.1.0PublishedPublic preview · 2026-11-02Release
v0.1.1PlannedAdded here once verified and published.
v0.2.0PlannedAdded here once verified and published.
Build log

Entries come from the repository log, which separates verified work from unverified work.

How it was built

  1. Repository created

    New independent repository: Apache-2.0 license, README, no functional code yet.

  2. First vertical slice

    API key in SecretStorage and a bounded command that sends a mission goal to Meta Model API, with unit tests.

  3. Local Glimmer tool-calling check

    Local Muse Glimmer tested with tool calls before building the Agent loop on it.

  4. Permission-gated Agent

    Agent slice with permission profiles, one-use approvals and fixes found while closing the slice.

  5. Sidebar and Settings rework

    Home, conversation and unified composer; conditional Settings and a corrected model catalog.

  6. v0.1.0 public preview

    First public release with a VSIX, 269 automated tests passing and 0 vulnerabilities in production dependencies.

HACKATHON_WORK.md · CHANGELOG.md

Demos

Only runs that can be checked against evidence in the repository.

See it work

Video to add

Install and first Agent run

Placeholder: add the link when the video is final.

Video to add

Approvals and the host record

Placeholder: add the link when the video is final.

Video to add

Mission run

Placeholder: shown only with results that were verified.

Origin

An independent project

RAME Code is an independent open-source project, started from scratch on 2026-10-31 for the Meta Global AI Developer Hackathon 2026, Indie Developer track. Author: Antoni Caimari Caldés.

Its public code and Git history begin in this repository. No earlier private code, services, data or results are included or required.

Its design draws conceptual inspiration from earlier RAME research. That work is documented separately and is not part of this project: ramesystem.com.