Derive vs OSCAR
A side-by-side look at the Derive and OSCAR MCP servers: transport, auth, tools, quality signals, and the config to install either one in Claude, Cursor, Windsurf or VS Code.
Verdict: choose Derive for developer tools work as a local process; choose OSCAR for developer tools as a local process. You can also run both side by side in one mcpServers block.
Which MCP server should you choose?
Choose Derive when
- You need dedicated capabilities in the Developer Tools domain.
- You prefer a local stdio subprocess transport.
- You have the keys it needs:
DERIVE_SERVER, DERIVE_TOKEN. - You want tools like
list_workspaces, list_artifacts, search.
Choose OSCAR when
- You need dedicated capabilities in the Developer Tools domain.
- You prefer a local stdio subprocess transport.
- Its auth model fits you: No auth required.
- You want tools like
status, list_projects, read_project.
Feature & specification comparison
Rows where the two servers differ are marked.
| Spec | ||
|---|---|---|
| Category | Developer Tools | Developer Tools |
| Quality signal | 64/100 · Good | 68/100 · Great |
| Install command | Confirmed | Confirmed |
| Transport | Local subprocess (stdio) | Local subprocess (stdio) |
| Auth | Not declared | No auth required |
| Pricing | Not declared | Free |
| Required env vars | DERIVE_SERVERDERIVE_TOKEN | None listed |
| Verified owner | Community listing | Community listing |
| GitHub stars | 30 | 192 |
| Directory activity | 0 views 0 installs 0 upvotes | 0 views 0 installs 0 upvotes |
| Tools | 8 | 11 |
| Full listing | View Derive | View OSCAR |
Tools & capabilities
Derive tools 8
list_workspacesList every workspace signed in on this machine you can act in — id, name, your role, local description, and which is active. One login reaches them all; pass a workspace's id or name as the `workspace` argument to list_artifacts / read / catch_up / comment / publish to act there for that call.list_artifactsList the artifacts in your workspace — short id, title, kind, current version, access, and browse `tags`. Defaults to this session's workspace; pass `workspace` (id or name from list_workspaces) to list another. Pass `tag` to list only artifacts carrying that tag (organize shows the vocabulary). Start here to find what to work on, then catch_up or read it.searchFind text within ONE artifact, or across a WORKSPACE — same tool, same behavior as the remote MCP server's `search`. Pass short_id to grep one artifact: matching lines with line numbers (and optional context), ripgrep-style, so you can then `read` a narrow `lines` range or `edit` that spot. Omit short_id to search across the workspace — the artifacts you can see, ranked by relevance and grouped by artifact — find WHICH doc has something before opening it. Searches the exact source by default (in:'text' searches the visible text instead). The query is matched literally (metacharacters are not special).readRead an artifact's CONTENT by short id, as Markdown by default (HTML is converted). Omit `section` to see the outline first (heading slugs for a single-file doc, page paths for a bundle) — call again with a `section` (or "*" for the full document) once you know what you want. Pass `format:'html'` for the exact source (needed before publish `edits`), or a past `version` for history. Also accepts derive://guide, /connect, or /compatibility so the onboarding strings in server instructions work even when MCP resources do not. For what CHANGED or the comment threads, use catch_up instead.catch_upSTART HERE on an artifact. Its state in one call: a summary, the review round, the versions since `since_version`, the open (and outdated) comment threads, and the full version history. Pass `comments` (open / addressed / resolved / outdated) to instead get that filtered thread list — your feedback queue. Pass `response_format='detailed'` (optionally with `since_version`/`to_version`) to fold in a line diff between two versions — of their readable Markdown form, not raw HTML. WAITING ON A REVIEW? Pass `wait` (seconds, max 50) to block until the human sends back — chain these instead of sleeping between polls.commentLeave feedback, reply in a thread, react, and/or resolve or reopen a thread. Anchor a NEW comment to text with `quote`, or to a linked-bundle loop/graph part with `visual_target`. Reply by passing the thread id as `reply_to`.organizeTags and collections in one tool — the library's findability layer. • READ (no `short_ids`): the workspace's tag vocabulary (tag → count) and its collections. Call this before tagging to reuse an existing tag over a near-duplicate. • READ (with `short_ids`): those artifacts' current tags + collections, plus `suggested` tags drawn from the most semantically-similar docs (when one id is given). • WRITE: pass `add`/`remove`/`set` to change tags, `collection` to file artifacts, or `state:'archived'`/`'live'` to archive and restore. Each artifact is authorized on its own; ones you can't touch are skipped. Tag freely and reuse the vocabulary — a well-tagged library is findable. Collections are heavier: a tag for plain findability, a collection when a set is a real unit.publishPublish a single-file artifact and get a permanent URL, versions, and feedback (a chat-local artifact has none of these). OMIT short_id to create a NEW artifact (title recommended); PASS short_id to publish a new version (same URL). Provide the body as `content_path` (a local file this server reads and uploads — preferred, zero tokens) or `content` (inline text). To CHANGE PART of an existing artifact, prefer `edits` (exact-match search/replace against the stored source — read format:'html' first) over resending everything. Pass `addresses` with the thread ids this revision resolves. (Multi-page bundles are published via the web app or the remote /mcp server.) FULLY-STYLED HTML renders as-authored (own <style>/scripts/fonts) in the sandboxed viewer — declare your own <meta name="viewport"> to skip the mobile-reflow injection, and self-host binaries via POST /v1/assets (images and woff2 fonts) instead of inlining base64.
OSCAR tools 11
statusOSCAR's version, its ports (HTTP, the fixed OSC listening port, the socket bridge), whether editing is locked, the MIDI driver's condition, and how many projects and published interfaces there are. Start here.list_projectsEvery project OSCAR holds, with its id and its title, as File > Open project or template lists them. A project is what a person edits; it reaches phones only once it is published (list_published).read_projectEvery widget in one project, published or not, with its full settings as the editor holds them: addresses, ports, channels, MIDI mappings, DMX targets. Also says where the project is published, if it is. A setting nobody touched reads as its default. What phones see is the published copy, which may be older: read_published reads that one.list_publishedEvery published interface: its name, the project it was published from, the path it is served at, who can open it (access: "network", or "off" for one no device can open, whose schedules and bridges still run), and how many widgets it holds. Published interfaces are running -- phones and tablets may be on them, and the rig hears them.read_publishedEvery widget on one published interface, with its full settings as phones have them: addresses, ports, channels, MIDI mappings, DMX targets. For answering why something does or does not respond.installation_mapOne picture of the whole installation: this computer's network addresses and ports, its MIDI ports and USB interfaces (a USB DMX interface among them), and everything the published interfaces talk to and listen for -- each OSC destination with the addresses sent to it, each MIDI port with its channels and controllers, each DMX output with its universe and channels -- with the control and interface behind every one. notices lists what does not add up: a MIDI port or USB interface that is not there, OSC followed while OSCAR cannot listen, two controls on the same DMX channels. It is worked out from settings and sends nothing onto the network: it says where OSCAR sends, not whether anything is there to hear it.midi_portsThe MIDI inputs and outputs this machine has right now, and the driver's condition.recent_activityThe network log's recent rows: what OSCAR's server actually received and sent, oldest first, repeats within a moment gathered into one row with a count n. Each row has what (OSC address, MIDI control, DMX universe and channels) and value (what the latest carried). Incoming rows: device (the sender's address, or the MIDI port), surfaces (published interfaces that follow it), canvas (the editor's canvas followed it); with neither, nothing follows it. Outgoing rows: to (where it went), origin (whose move: canvas, local, internet, schedule, bridge), surface (the published interface) and widget. dropped, on either, says why it was refused. The first place to look when something moves that should not, or does not move when it should.describe_widgetsEvery widget OSCAR has -- tag, attributes, defaults, and each setting with its type, options and meaning -- plus how an interface is written as HTML. Read this before building one.validate_draftRun an interface written as HTML through the widgets' own validators: unknown settings, values their panels would refuse, missing or repeated ids. Returns the complaints to correct. Nothing is saved.create_draftValidate an interface and, if it is clean, save it as a draft: it appears in OSCAR under File > Open project or template, with a Draft badge. It is a file for a person to review: nothing is published, nothing is sent, and no project is changed. An existing draft of the same name is only replaced when overwrite is true.
Ready-to-paste client configs
Paste either block (or both) into your client config, e.g. claude_desktop_config.json or ~/.cursor/mcp.json. We only show a config when the install command is confirmed.
Derive
{
"mcpServers": {
"derive": {
"command": "npx",
"args": [
"-y",
"@derive-to/mcp"
],
"env": {
"DERIVE_SERVER": "YOUR_DERIVE_SERVER_HERE",
"DERIVE_TOKEN": "YOUR_DERIVE_TOKEN_HERE"
}
}
}
}
OSCAR
{
"mcpServers": {
"oscar": {
"command": "npx",
"args": [
"-y",
"createwithoscar"
]
}
}
}
Frequently asked questions
Derive is categorized under Developer Tools and uses a local stdio subprocess. OSCAR belongs to Developer Tools and uses a local stdio subprocess. Pick Derive for developer tools work and OSCAR for developer tools.
Copy the mcpServers block from the config section on this page into your client's config (for example claude_desktop_config.json or ~/.cursor/mcp.json), then fully restart the client.
Derive has no declared pricing — check the project's own site. OSCAR is listed as free. Auth: no auth required. If a server needs environment variables such as API keys, define them under the env key in your client's MCP config.
Yes. MCP clients support multiple servers: add derive and oscar as separate keys in the same mcpServers object, and your assistant will call tools from either one.
On AllMCPs, Derive has a quality signal of 64/100 (Good) with 0 views, 0 installs and 30 GitHub stars. OSCAR has 68/100 (Great) with 0 views, 0 installs and 192 GitHub stars.