Skip to main content
Glama

Server Configuration

Describes the environment variables required to run the server.

NameRequiredDescriptionDefault
CNB_TOKENYesCNB personal access token (PAT) used by the server to call CNB; keep it server-side only.
MCP_WRITE_MODENoDefaults to `read-only`; set to `write` to enable write tools.read-only
MCP_ACCESS_TOKENSYesClient Bearer token(s) for MCP access. OAuth authorization uses this to verify the server owner. Generate an independent secret, e.g. with `openssl rand -hex 32`.
MCP_OAUTH_CLIENT_IDNoPreconfigured OAuth client ID for ChatGPT. Leave empty to disable OAuth.
MCP_PUBLIC_BASE_URLNoPublic HTTPS origin URL, for example `https://mcp.example.com`.
MCP_OAUTH_CLIENT_SECRETNoPreconfigured OAuth client secret for ChatGPT. Leave empty to disable OAuth.
MCP_OAUTH_REDIRECT_URISNoExact HTTPS callback URIs shown in the ChatGPT management page, separated by commas.

Instructions

Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.

Unofficial remote MCP server for CNB.cool. Read-only by default; write tools appear only when the deployment enables write mode. Destructive tools require an explicit confirm=true. Build/NPC SNs returned by trigger tools must be polled via build_status / npc_status — a successful trigger means accepted, not finished.

Capabilities

Features and capabilities supported by this server

Protocol revision2025-11-25

CapabilityDetails
tools
{
  "listChanged": true
}

Tools

Functions exposed to the LLM to take actions

NameDescription
whoamiA

Return the CNB user that the upstream token belongs to. Use this first to verify authentication works.

repo_getB

Get metadata of a CNB repository (description, visibility, default branch, stats).

repo_listB

List repositories owned by the authenticated CNB user.

repo_searchB

Full-text search across public CNB repositories.

repo_file_readA

Read the raw text content of a file at a given ref (branch/tag/commit).

repo_tree_listB

List files and directories at a path. Omit path to list the repository root.

repo_commits_listB

List commit history, optionally starting from a commit-ish.

repo_commit_getB

Get one commit with its changed files (patch).

repo_diff_compareC

Get the diff between two refs (base and head), e.g. base=main head=feature-x.

repo_commit_statusesB

Get CI check statuses (pipeline results) attached to a commit.

repo_branch_listC

List branches of a repository.

repo_branch_getB

Get one branch with its head commit.

repo_tag_listC

List tags of a repository.

issue_listC

List issues of a repository with optional filters.

issue_getC

Get one issue by number.

issue_comment_listC

List comments of an issue.

pr_listB

List pull requests with optional filters.

pr_getB

Get one pull request including mergeable_state (conflict/merge status) and CI statuses.

pr_filesB

List files changed by a pull request with patch diffs.

pr_commitsC

List commits contained in a pull request.

pr_checksB

Get CI check statuses attached to a pull request.

pr_comment_listC

List comments on a pull request.

pr_review_listA

List reviews of a pull request (approve / request changes / comment).

build_listC

List CI build history of a repository with filters.

build_statusA

Get the overall status and per-pipeline status of one build (by SN, e.g. cnb-xxx). SN values come from build_list / build_trigger / npc_dispatch.

build_stage_logsA

Get the log of one pipeline stage. Find pipelineId/stageId via build_status first. Log output is truncated to keep responses small.

npc_searchC

Search NPC (automated agent) characters available on CNB.

npc_statusA

Get the real status of a dispatched NPC task (build). Returns overall status, per-pipeline status and the NPC that triggered it. Use build_stage_logs for detailed logs.

npc_actionsB

List recorded NPC actions (calls, pushes, PRs) in a repository.

Prompts

Interactive templates invoked by user choice

NameDescription

No prompts

Resources

Contextual data attached and managed by the client

NameDescription

No resources

TDQS

C2.9/5.0

Scored across 29 tools

Disambiguation4/5

Tools are grouped by resource prefix (repo_*, issue_*, pr_*, build_*, npc_*) and most have a single distinct action, so an agent can pick correctly in the common case. A few pairs are close: repo_commit_statuses vs pr_checks, build_status vs npc_status, and pr_commits vs pr_files all surface overlapping CI/commit data, though the descriptions do enough to separate them.

Naming Consistency4/5

Nearly all tools follow a resource_action snake_case convention (repo_list, issue_get, pr_review_list, build_stage_logs), which is highly predictable. Minor deviations exist: whoami and pr_commits (verb/plural noun without action) and a couple of prefix irregularities like repo_file_read vs repo_tree_list.

Tool Count2/5

At 29 tools this is above the comfortable range for what is essentially a read-only repository browsing surface, and several list/get pairs feel granular enough to be merged. The npc_* group in particular adds scope that inflates the set without adding much distinct capability.

Completeness3/5

Read coverage is thorough: repos, files, commits, diffs, branches, tags, issues, PRs, reviews, builds, and logs are all reachable. However, there is no write/lifecycle operation anywhere — no create issue, open PR, comment, merge, or branch creation — so workflows dead-end at read-only inspection.

Maintenance

ActivityNo data
ResponsivenessOutdated data