Carbon Forge MCP Server

by 100monkeys-ai Your server? Claim it

answering

Carbon Forge is answering right now. Last checked 2 min ago. It exposes 22 tools.

One record of code, files and pages that people and agents read and write, with lanes and git.

Uptime history 47 hours of history · worst hour 43%

47 hours agonow

50.3%

Uptime 24h

92 of 183 checks

Tools

read from the server

1242 ms

Response time

average over 24h

open, no key

Access

streamable-http

Carbon Forge does not always answer

Over the last week it answered 50.0% of our checks. We check every 15 minutes, so you hear about the next outage within the hour — not from your users.

Connect this server

Endpoint below is the one we actually reach during checks — not the one copied from a README. Last verified 2 min ago.

Claude Code Claude Desktop Codex Cursor VS Code

run in your terminal

claude mcp add carbon-forge --transport http https://carbonforge.dev/api/mcp

Settings → Connectors → Add custom connector, then paste this address

https://carbonforge.dev/api/mcp

~/.codex/config.toml

[mcp_servers.carbon-forge]
url = "https://carbonforge.dev/api/mcp"

.cursor/mcp.json

{
  "mcpServers": {
    "carbon-forge": {
      "url": "https://carbonforge.dev/api/mcp"
    }
  }
}

.vscode/mcp.json

{
  "servers": {
    "carbon-forge": {
      "type": "http",
      "url": "https://carbonforge.dev/api/mcp"
    }
  }
}

This server publishes 1 more address. The block above uses the one we reach during checks; the full list is under Endpoints below, and the author may intend a particular one for your client.

Available tools 22

Read directly from the server with tools/list, grouped by what they act on. If a tool disappears, we record the date.

discovery

discovery.list_entities

Browse pages or atoms inside a workspace with faceted filters (tag, author, date range, visibility, metadata containment) and sort by alphabetical, recency, or star count. Viewer-aware: members see everything, anonymous callers see only public entities in public workspaces. 50 items per page, cursor-based. Details: https://carbonforge.dev/docs/p/api/mcp-tools-catalog

discovery.list_instances

List instances. Filter: all (default) | public | my | my_admin. public lists an instance that has public reading on, is listed in Discover, is not frozen, and has a workspace that is not private holding a page that is not private. my and my_admin list the caller's own instances (member; admin) with no such gate; all is public plus my. Without a token every filter is public.

discovery.list_workspaces

List workspaces across one or all instances, with the same filters. public lists a workspace when its instance has public reading on, the workspace is not private, its instance is listed in Discover and not frozen, and it holds a page that is not private. my and my_admin list the caller's own workspaces with no such gate; all is public plus my. Without a token every filter is public.

discovery.search

Search instance and workspace metadata + grounding documents. Returns parallel lists for both scopes.

revisions

revisions.compare

Markdown-level diff between two revisions of the same page or atom. Returns segments (op: equal | insert | delete) plus both Markdown projections.

revisions.get

Fetch a specific revision (returns Markdown body).

revisions.list

List revisions for a page or atom (cursor-paginated).

atoms

atoms.list

List atoms in the current workspace. The `path_prefix`, `body_contains`, `metadata`, `sort`, `direction` and `include_metadata` arguments mirror the same arguments on pages.list.

atoms.read

Read an atom by path or UUID with usage_pages projection. Returns metadata. Pass `section` with a heading's plain text to get back only that heading and the blocks under it, down to the next heading at the same or shallower level; the result then also carries `section: { heading, level, matches }`, and everything else on it is the atom's own, so a section can be read and written straight back.

pages

pages.list

List pages in a workspace. `path_prefix` narrows by leading segment; `body_contains` keeps pages whose Markdown holds a literal substring; `metadata` keeps pages whose top-level metadata contains the given scalar keys; `sort` (path | updated_at | created_at) and `direction` order the set; `include_metadata` adds each page's metadata. Details: https://carbonforge.dev/docs/p/api/mcp-tools-catalog

pages.read

Read a page by path or UUID. Returns the Markdown body plus the page metadata object. Pass `section` with a heading's plain text to get back only that heading and the blocks under it, down to the next heading at the same or shallower level; the result then also carries `section: { heading, level, matches }`. Details: https://carbonforge.dev/docs/p/api/mcp-tools-catalog

workspaces

workspaces.list

List every workspace the calling user is a member of. This is a membership catalogue and not a capability: it is not narrowed by the token, so a listed workspace may still be outside this token's reach. Call cortex.ground for what this token can actually resolve.

workspaces.resolve_slug

Resolve a workspace slug to its id, role, and archived state. 404 when the caller isn't a member (no enumeration).

comments

comments.list

List the comment thread for a page or atom.

cortex

cortex.ground

Orientation surface: instance metadata, both grounding documents, identity and capabilities, reachable workspaces, sidecar health. Pass `workspace` to ground a named workspace without moving the per-token pointer. Pass a `known_grounding_at` / `known_workspace_grounding_at` you hold and an unchanged document is acknowledged, not resent. Details: https://carbonforge.dev/docs/p/api/mcp-tools-catalog

cross

kg.list_cross_links

List incoming + outgoing cross-links for a page or atom.

files

files.list

Browse the workspace as a file tree, one level at a time: directories, pages, atoms, code trees (expanded at their paths) and library entries. Each node names its kind, entity, current revision and digest; `at` holds the entities exactly at `path`. `surface` keeps that surface only. History: files.history. Anonymous callers see public pages and atoms, and code trees of surface code.

global

search.global

Hybrid FTS + pgvector + RRF search. Returns Markdown excerpts.

literal

search.literal

Exact-substring search over the Markdown projection of pages and atoms — deterministic, unlike hybrid `search.global`. A needle shorter than 3 characters is rejected. Returns one envelope per matched entity, up to 5 hits each as `{ line, col, snippet }`. `case_sensitive` defaults to false; `scope` is workspace | all_mine | all_public. Details: https://carbonforge.dev/docs/p/api/mcp-tools-catalog

media

media.read

Fetch metadata for a single media item.

kg.get_related

Semantic nearest-neighbors via similarity_edges.

trees

trees.get

Read a code tree's revision (the current one by default), its metadata and surface, and a page of its entries in path order: path, git mode, size and blob SHA-256, optionally under `path_prefix`. A symlink's blob is its target; a submodule link has no blob and gives its `commit`. With `lane`, `metadata` is the lane's bag if it sets one. Anonymous callers read a public code tree's current one.

Endpoints

URL	Transport	State	Latency	Checked
https://carbonforge.dev/api/mcp	streamable-http	answering	1473 ms	2 min ago
https://{instance}.carbonforge.dev/api/mcp	streamable-http	answering	0 ms	5 min ago

Alternatives to Carbon Forge

same job, measured the same way

Tenki Sandbox MCP

by luxorlabs

Disposable microVM sandboxes for AI agents: run code, read/write files, git, preview URLs.

86 installs/wk local only

cpcp

by builderbasetech-lgtm

Send files and text with a six-digit transfer code. Remote MCP for agents.

answering

Kin

by kinlab

A graph-native code repository for AI agents and the people who build with them.

156 installs/wk local only

Repomix

by yamadashy

Pack local or remote codebases into AI-friendly files that LLMs and coding agents can read or search

58 596 installs/wk local only

MCP Github

by infoinlet-marketplace

GitHub for AI agents — repos, files, code search, issues, PRs, commits. Read-only default.

9 installs/wk local only

DESIGN.md Linter & Extractor

by openkrill

Lint and diff DESIGN.md files with the official Google linter, or write one from a web page.

6 tools answering

Agent Billboard

by yourmatematt

Read and post to The Agent Billboard on Solana, with spend limits in code and a reasoning log.

56 installs/wk local only

Graphicle

by graphicle-ai

Code graph and project memory for coding agents, cited to exact files and lines, on your machine.

local only

Carbon Forge — questions

Answers built from our own checks of this server.

What can Carbon Forge do?

It exposes 22 tools, read directly from the server on our last check. Among them: atoms.list, atoms.read, comments.list, cortex.ground, discovery.list_entities, discovery.list_instances and 16 more. The full list with descriptions is on this page — we take it from the server itself via tools/list, not from a README. How MCP servers expose tools in the first place →

What is Carbon Forge mostly used for?

Its tools cluster around discovery, revisions and atoms. That is what this server is built to work with — the grouping comes from the actual tool names, not from a category we assigned.

Is Carbon Forge working right now?

We send a real MCP handshake every 15 minutes. Over the last 24 hours 92 of 183 checks got a reply (50.3%), average response time 1242 ms. The bar chart above shows every period we have measured.

How do I connect Carbon Forge?

Copy the ready config from this page. We generate it for Claude Code, Claude Desktop, Codex, Cursor and VS Code, each in the form that client actually reads. It is a remote server, so there is nothing to install — the client connects to the address.

Does Carbon Forge need an API key?

No. Carbon Forge completed a full MCP handshake with us as an anonymous client and listed its tools without asking for anything. All 22 of them are readable on this page. This is what we observed, not what the docs claim.

How fast is Carbon Forge?

It answers our handshake in 1242 ms on average, which is faster than 2% of all working MCP servers we measure. That is on the slow side — worth knowing if the tool sits inside an interactive loop. The comparison comes from our own checks across the whole registry, every 15 minutes.