ChatGPT Plugin

Ronda

Edit your web app from chat

Brand: ronda

Category: Pending
Primary Subcategory: Pending

Integration details

Description

Edit existing web apps from ChatGPT. Connect a GitHub repository and ronda starts an isolated copy of the app with a running dev server and a live preview link. ChatGPT reads the code, makes small changes, checks the build output and shows you the result in the preview. Changes ChatGPT makes directly stay in the sandbox until you ask for a pull request. Changes made by ronda's hosted agent are pushed to a working branch in your repository (its name starts with ronda/session-), never to your default branch. Use it to change page copy, fix layout bugs, adjust styles or update components without setting up a local environment. ronda works with repositories you connect through your own GitHub account and only acts on projects in your ronda workspace. Limitations: the app must be able to run as a web project in a single container, setup can take a minute or two, and scripts run in the sandbox have time and output limits.

Integration type: Plugin
Verification status: Not applicable
Platform: ChatGPT
Category: Pending
Primary Subcategory: Pending
Secondary Subcategories: None listed
Brand: ronda
Access: Account required
First tracked: 2026-10-17
Tool count: 25
Geography: US

Agent DiscoverabilityChatGPT Plugin Discovery Score

ChatGPT organic discovery is live

Your score is coming

ChatGPT now suggests Plugins on its own when they match a user's request.Your Plugin Discovery Score measures how often yours appears, and it will show here as soon as it’s ready.

ChatGPT recommends Airtable when a user asks it to update a CRM. — ChatGPT surfaces Airtable because it matches the user's request.

Primary Subcategory

Competitive lineup

No Primary Subcategory assigned yetOnce Ronda has a Primary Subcategory, the Plugins and Connectors it competes with will appear here with their scores.

Tools

25 tools agents can invoke

Add npm registry packages (1-10, e.g. 'zod' or '@scope/name@^1.2.0') to the project with its package manager, then commit the package.json and lockfile changes on the session's ronda branch. Registry specs only: URLs, git or file specs and flags are rejected. Contacts the package registry, and packages' own install scripts run in the sandbox. Can replace installed versions and rewrites the lockfile; repeating the same specs usually changes nothing, but a range or tag can resolve to a newer version. Timeout default 300s.
install_package
Connect a GitHub repository (owner + name) to ronda as a project so it can be opened in a sandbox. Auto-detects framework, package manager, dev command and port; pass devCommand only if the user confirms an override. Creates a project record in the workspace and does not modify the repository. Idempotent: an already-connected repo returns its existing projectId (alreadyExisted:true). Next: open_session.
create_project
Create a NEW file in the sandbox repository and commit it on the session's ronda branch (not pushed to GitHub until open_pr). Never overwrites: fails with FILE_ALREADY_EXISTS if anything exists at the path. Parent directories are created. Use apply_patch to change an existing file.
create_file
Ask ronda's hosted coding agent to make a change from a natural-language instruction. Runs on the workspace's own model key (returns a setup link if none is configured) and commits on the session branch. Returns an editId immediately; poll get_edit. Pass idempotencyKey to make retries safe for 24h. Prefer editing directly with apply_patch when you can. The agent may rewrite or delete existing files and sends code to the workspace's configured model provider.
instruct
Permanently tear down a sandbox session and its container. Any edits not yet in a pull request are lost. Idempotent (already-destroyed returns success). Only call when the user asks.
destroy_session
Change existing files in the sandbox repository and commit it on the session's ronda branch (not pushed to GitHub until open_pr). The preferred way to edit. Accepts a unified diff or a '*** Begin Patch' envelope; it can modify, rename and delete lines and files. All-or-nothing: on mismatch nothing changes and the error explains why (re-read the file and regenerate). Retrying an already-applied patch returns alreadyApplied:true. The preview hot-reloads. Changes take effect in the isolated sandbox, where the project's code (including package.json scripts you change) can run via run_script.
apply_patch
Have ronda's hosted agent produce 2-3 alternative versions of one change (each committed on its own branch) and restart the sandbox with a variant switcher in the preview so the user can compare. Uses the workspace's model key. Synchronous, 30-90s. The agent rewrites code and sends it to the workspace's model provider. Resolve with resolve_variant.
request_variants
Show everything changed in this session compared with where it started: per-file summary plus the unified diff (committed and uncommitted). Use before summarizing changes or opening a PR. Read-only.
get_diff
Return the ronda account and workspace this connection acts as (opaque stable id, email, workspace name). Read-only.
get_profile
Return the latest dev-server output (compile errors, runtime errors) or the dependency-install log. Use after an edit to check the app still builds. Read-only.
get_logs
Poll an instruct edit: queued | in_progress | applied | failed | cancelled. When applied, includes summary, files changed and previewUrl. Read-only.
get_edit
Get a sandbox session: sandboxStatus (CREATING → INSTALLING → STARTING → READY), branch, previewUrl once READY, and recent edits. Use to poll after open_session and to fetch the preview link. Read-only.
get_session
Keep the chosen variant and discard the others (removes the variant switcher from the code; discarded variants can't be recovered from the preview). Uses the workspace's model key; synchronous, 30-60s; restarts the sandbox.
resolve_variant
List GitHub repositories the user's connected GitHub account can access (most recently pushed first, max 100): owner, name, fullName, defaultBranch. Use to resolve a repo the user names (e.g. github.com/acme/site) before create_project. If GitHub isn't connected, returns a link the user can open to connect it. Read-only.
list_repos
List files and directories in the sandbox repository (relative paths, sizes). Skips .git, node_modules and build output; max 500 entries. Read-only.
list_files
List a project's sandbox sessions (newest first, max 50), optionally including destroyed ones, plus the plan's concurrent-sandbox cap. Read-only.
list_sessions
List the web app projects (connected GitHub repositories) in the user's ronda workspace, newest first, max 100. Returns id, name, GitHub repo, dev command/port and primary service. Use first to find which project the user means; if their repo isn't listed, use list_repos then create_project. Read-only.
list_projects
Start (or reuse) an isolated sandbox for a project: clones the repo on a fresh ronda branch, installs dependencies and runs the dev server with a live preview URL. Reuses a running sandbox or resumes an archived one when available. Returns immediately; poll get_session until sandboxStatus is READY (typically 30-120s). Consumes the workspace's sandbox hours. Does not change the GitHub repository.
open_session
Push the session branch to the user's GitHub repository and open a pull request against the default branch. Commits any uncommitted sandbox changes first. Visible to everyone with access to the repo — only call when the user explicitly asks to open a PR. Title/body default to the edit summaries. Calling again after success opens no duplicate; it rotates to a new branch only if a PR already exists for it.
open_pr
Read a text file from the sandbox repository, optionally a line range (max 2000 lines per call; use startLine/endLine to page). Environment files (.env*) are not returned. Read-only.
read_file
Write a file's ENTIRE content, replacing whatever is there (or creating it), and commit it on the session's ronda branch (not pushed to GitHub until open_pr). Overwrites existing content, so prefer apply_patch for edits and create_file for new files; use this only to rewrite a file wholesale. Writing identical content is a no-op. Changes take effect in the isolated sandbox, where the project's code (including package.json scripts you change) can run via run_script.
write_file
Restart the sandbox dev server (e.g. after installing dependencies or changing config) and wait up to 60s for it to come back. The preview is briefly unavailable and in-flight work in the dev server (running builds, requests, unsaved dev-server state) is cancelled. Does not change files.
restart_dev_server
Run one named script from package.json (e.g. build, lint, typecheck, test) with the repo's package manager (pnpm, yarn or npm, from the lockfile) and return exit code, stdout and stderr (last 64 KB each). This executes the project's own code: the script is whatever package.json defines at that moment, including any change made with apply_patch or write_file. Only existing scripts run; an unknown name lists the available ones. Optional args are passed as plain argv, never through a shell. Scripts can modify sandbox files and reach the internet; changes are not committed automatically. Hard timeout (default 120s, max 600s). Not idempotent.
run_script
Search the sandbox repository for text (literal by default, or regex) and return matching file, line number and line text. Respects .gitignore, skips .env files. Use to find where copy, components or styles live. Read-only.
search_code
Apply a previously proposed A/B experiment: writes the variant code and creates the feature flag and experiment in the workspace's connected PostHog project, stopping before launch (the user launches it in PostHog). Uses the workspace's model key. Requires a PostHog connection. The agent rewrites code and creates the flag and experiment in PostHog (not launched).
apply_experiment

FAQs

Ronda ChatGPT Plugin FAQ

How the directory, categories and Discoverability Score work.

Read the methodology

How do I improve Ronda's ChatGPT Plugin discoverability?

The levers are the listing surface agents actually read: names, descriptions, keywords, tool metadata, and registry health. Which lever matters depends on where discovery breaks, which is what continuous measurement shows.

Where is this profile measured?

This profile uses the geography attached to the latest public registry snapshot: US. Locale tags are intentionally omitted.