Agent Token Budget MCP
Agent Token Budget MCP is an MCP server that enforces per-session token spend caps and returns signed attestations when a budget is exhausted.
About
Agent Token Budget MCP is a stdio MCP server published on PyPI that lets developers and small teams put a hard per-session ceiling on model token spend. Instead of discovering runaway costs after the fact, the agent runtime can consult the server as usage accumulates and stop or redirect work when the cap is reached. When the budget is exhausted, the server issues signed attestations you can log, show in support threads, or feed into other automation so "why did this stop?" has a verifiable answer. It targets developers shipping agent features, internal copilots, or API wrappers who need lightweight guardrails without building a full billing platform. Install the package, wire it into your MCP client, set session budgets, and treat it as operational cost control alongside your chosen LLM provider—not a replacement for provider dashboards or enterprise FinOps suites.
- Per-session token spend cap enforced by the MCP server
- Signed budget-exhausted attestations for audit and downstream gates
- stdio transport via PyPI package agent-token-budget-mcp v1.1.1
- Pairs conceptually with broader budget/fleet tooling (twin of bf)
- Runs as a local MCP process for Claude Code, Cursor, and similar hosts
Agent Token Budget MCP by the numbers
- Data as of Aug 10, 2026 (Skillselion catalog sync)
claude mcp add agent-token-budget-mcp -- uvx agent-token-budget-mcpAdd your badge
Show developers this MCP server is listed on Skillselion. Paste this into your README.
| Package | agent-token-budget-mcp |
|---|---|
| Transport | STDIO |
| Auth | None |
What it does
Cap how much each agent session can spend on tokens and get a signed record when the budget is hit.
Who is it for?
Best when you're running Claude Code or Cursor agents against paid APIs and want session-level spend caps and attestations without a full billing stack.
Skip if: Skip if you only need monthly invoices from OpenAI or Anthropic with no per-session enforcement or signed stop events.
What you get
Each session runs under a defined token budget, and when it hits the cap you get signed budget-exhausted attestations you can audit or chain into other tools.
- Enforced per-session token spend ceiling during agent runs
- Signed budget-exhausted attestations when a session hits its cap
- stdio MCP integration listed as agent-token-budget-mcp v1.1.1
By the numbers
- MCP server version 1.1.1
- 1 PyPI package with stdio transport
- Registry identifier: agent-token-budget-mcp
Recommended MCP Servers
Agent Audit MCPImmutable audit trail for agent-to-agent interactions2
Agentguard47 MCPRead-only MCP server for coding-agent traces, alerts, costs, usage, and budget health.4
Agentic Observability MCPAgent tracing, cost tracking, anomaly detection for LLM agentsAgent Incident Relay MCPAgent Incident Relay MCP — Article 73 5-clock broadcaster.
Agent Infra MCP18 tools: agent memory, API drift detection, and HMAC-signed trust receipts
Agentmonitorrelay MCPAI agent run monitoring with incident replay and SLA receipts.How it compares
MCP spend-guard integration, not an LLM skill or provider billing portal.
FAQ
Who is agent-token-budget-mcp for?
It is for and developers who run coding agents on metered LLM APIs and need per-session caps plus signed proof when a budget runs out.
When should I use agent-token-budget-mcp?
Use it in production or heavy local agent workflows once token cost becomes a risk—before unattended loops, demos, or customer-facing agent features go live.
How do I add agent-token-budget-mcp to my agent?
Install the PyPI package agent-token-budget-mcp, add it as a stdio MCP server in Claude Code, Cursor, or another MCP host, and configure per-session budget settings in that host’s MCP config.