Schema | teamspend

teamspend

Overview Schema Source Related Servers Discussions Admin

Server Configuration

Describes the environment variables required to run the server.

Name	Required	Description
`CODEX_HOME`	No	Overrides the Codex CLI home directory (default ~/.codex).
`XDG_CONFIG_HOME`	No	Overrides the base config directory for Claude Code (and other tools).
`CLAUDE_CONFIG_DIR`	No	Overrides the Claude Code configuration directory (default ~/.claude).
`OPENCODE_DATA_DIR`	No	Overrides the OpenCode local data directory (default ~/.local/share/opencode/storage/).
`TEAMSPEND_COPILOT_ORG`	No	Your GitHub org login.
`TEAMSPEND_CURSOR_TOKEN`	No	Cursor Admin API key.
`TEAMSPEND_COPILOT_TOKEN`	No	GitHub token with read:org scope or fine-grained 'View Organization Copilot Metrics' permission.
`TEAMSPEND_CLAUDE_CODE_TOKEN`	No	Anthropic Admin/Analytics API key.
`TEAMSPEND_COPILOT_SEAT_PRICE_USD`	No	Optional flat seat price per user for Copilot; if set, added once per active user for the whole comparison window.

Instructions

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

This server publishes no instructions, or was last inspected before Glama recorded them.

Capabilities

Features and capabilities supported by this server

Protocol revision2025-11-25

Capability	Details
`tools`	{ "listChanged": false }
`prompts`	{ "listChanged": false }
`resources`	{ "subscribe": false, "listChanged": false }
`experimental`	{}

Tools

Functions exposed to the LLM to take actions

Name Description

Name	Description
runA	Compare AI-coding-tool spend between a "before" and "after" date window across two tools, using each vendor's own admin/analytics API (or local session logs for the credential-free `claude-code-personal` mode) so the result is a real reported number, not an estimate. Call this when a user asks things like "did switching from Cursor to Claude Code save us money" or "what did our AI coding spend look like before vs. after we rolled out Copilot" -- it answers exactly that comparison in one shot. Do not call it for anything outside a two-tool before/after cost comparison (e.g. general usage analytics, per-line-item billing detail, or tools other than cursor, claude-code, claude-code-personal, copilot, opencode, codex). Prerequisites: the two admin API tokens for whichever tools are being compared must already be set as environment variables in the server's process (e.g. TEAMSPEND_CURSOR_TOKEN, TEAMSPEND_CLAUDE_CODE_TOKEN, TEAMSPEND_COPILOT_TOKEN + TEAMSPEND_COPILOT_ORG); claude-code-personal needs no token since it reads local session logs instead. Missing credentials do not raise an exception -- the tool still returns successfully, with a `DataUnavailableError` recorded against the affected side (see below) so the caller can see exactly what's missing. Behavioral notes: this makes outbound network calls to the relevant vendor APIs (skipped for claude-code-personal) and is read-only against the user's own systems, but it also writes a `teamspend-snapshot-*.json` report file to disk in the current working directory on every run (contains per-user email and spend data, permissions set to owner-only 0600). Each call is independent and idempotent -- rerunning with the same args re-fetches fresh data and overwrites the report file, it never accumulates state across calls. On a CLI crash or non-zero exit, the returned dict carries an `error` key plus the captured `stderr` and the exact `command` that was run, rather than raising. Parameter: `args` is a list[str] of argv exactly as they'd appear on the teamspend command line -- `--json` is appended automatically and must not be included. Required flags are `--tools <a>,<b>` (exactly two of: cursor, claude-code, copilot, opencode, claude-code-personal, codex) and `--before`/`--after` as `YYYY-MM-DD:YYYY-MM-DD` ranges. Optional flags: `--before-csv`/`--after-csv <path>` to backfill a window an API can't reach, `--breakdown session` for a per-session cost breakdown. Real examples: ["--tools", "cursor,claude-code", "--before", "2026-04-01:2026-04-30", "--after", "2026-06-01:2026-06-30"] ["--tools", "claude-code-personal,opencode", "--before", "2026-01-01:2026-01-31", "--after", "2026-02-01:2026-02-28", "--breakdown", "session"] ["--help"] (discovers every supported flag and tool name directly from the CLI, without a code change) Returns a dict parsed from the CLI's own `--json` output with keys `before` and `after` (each `{label, tool, result, error}`, where `result` is null and `error` is populated when that side's data was unavailable), plus top-level `deltaUsd`, `deltaPercent`, and `topSpendersAcrossBoth`.

runA

Compare AI-coding-tool spend between a "before" and "after" date window across two tools, using each vendor's own admin/analytics API (or local session logs for the credential-free claude-code-personal mode) so the result is a real reported number, not an estimate. Call this when a user asks things like "did switching from Cursor to Claude Code save us money" or "what did our AI coding spend look like before vs. after we rolled out Copilot" -- it answers exactly that comparison in one shot. Do not call it for anything outside a two-tool before/after cost comparison (e.g. general usage analytics, per-line-item billing detail, or tools other than cursor, claude-code, claude-code-personal, copilot, opencode, codex).

Prerequisites: the two admin API tokens for whichever tools are being compared must already be set as environment variables in the server's process (e.g. TEAMSPEND_CURSOR_TOKEN, TEAMSPEND_CLAUDE_CODE_TOKEN, TEAMSPEND_COPILOT_TOKEN + TEAMSPEND_COPILOT_ORG); claude-code-personal needs no token since it reads local session logs instead. Missing credentials do not raise an exception -- the tool still returns successfully, with a DataUnavailableError recorded against the affected side (see below) so the caller can see exactly what's missing.

Behavioral notes: this makes outbound network calls to the relevant vendor APIs (skipped for claude-code-personal) and is read-only against the user's own systems, but it also writes a teamspend-snapshot-*.json report file to disk in the current working directory on every run (contains per-user email and spend data, permissions set to owner-only 0600). Each call is independent and idempotent -- rerunning with the same args re-fetches fresh data and overwrites the report file, it never accumulates state across calls. On a CLI crash or non-zero exit, the returned dict carries an error key plus the captured stderr and the exact command that was run, rather than raising.

Parameter: args is a list[str] of argv exactly as they'd appear on the teamspend command line -- --json is appended automatically and must not be included. Required flags are --tools <a>,<b> (exactly two of: cursor, claude-code, copilot, opencode, claude-code-personal, codex) and --before/--after as YYYY-MM-DD:YYYY-MM-DD ranges. Optional flags: --before-csv/--after-csv <path> to backfill a window an API can't reach, --breakdown session for a per-session cost breakdown. Real examples: ["--tools", "cursor,claude-code", "--before", "2026-04-01:2026-04-30", "--after", "2026-06-01:2026-06-30"] ["--tools", "claude-code-personal,opencode", "--before", "2026-01-01:2026-01-31", "--after", "2026-02-01:2026-02-28", "--breakdown", "session"] ["--help"] (discovers every supported flag and tool name directly from the CLI, without a code change)

Returns a dict parsed from the CLI's own --json output with keys before and after (each {label, tool, result, error}, where result is null and error is populated when that side's data was unavailable), plus top-level deltaUsd, deltaPercent, and topSpendersAcrossBoth.

Prompts

Interactive templates invoked by user choice

Name	Description
No prompts

Resources

Contextual data attached and managed by the client

Name	Description
No resources

Deploy Server

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

TDQS

A4.6/5.0

Scored 2026-08-12 04:15 across 1 tool

Disambiguation5/5

With only a single tool, there is no possibility of confusing it with another. The tool's description clearly limits its purpose to a specific before/after cost comparison, so an agent cannot misselect it for anything else.

Naming Consistency4/5

There is only one tool named 'run', so there is no inconsistency in naming conventions. However, the name is very generic and does not convey the tool's specific function, which is a minor weakness since a descriptive name would be more helpful.

Tool Count3/5

The server exposes only one tool, which is below the typical 3-15 range. While the tool is substantial and covers a focused use case, the single-tool surface feels thin for a server, though it is not an extreme mismatch because the tool is not trivial.

Completeness4/5

Within its declared scope of comparing AI-coding-tool spend between two windows, the tool provides a complete workflow including error handling for missing credentials and a report file. Minor gaps like lacking multi-tool comparison or raw spend analytics don't affect its stated purpose.

Maintenance

ActivityMaintained

ResponsivenessNo issues

Resources

Need Help?

Related Servers