long-form

Claude Code was using 51,000 tokens before I even typed a prompt — I fixed it

Mahnoor Faisal opened a new Claude Code session, ran /context, and found 51,400 tokens already loaded. Disabling four test plugins and auto-memory got the starting context down to roughly 41,400 before any real prompt.

Published 2026-08-28Source: XDA
XDA source artwork

Why it matters

Starting context is a fixed overhead that gets reprocessed on every later turn, so a 10,000-token trim compounds over a long session. Most of the removable bloat came from add-ons installed once to try out and never cleaned up.

Tokenmaxxing read

Built-in system tools alone took 28,500 tokens, and no one can remove those. The controllable slice was plugins injecting skills, agents and session messages, plus auto-memory adding about 4,000 tokens to the system prompt. Audit agent add-ons the way you audit dependencies.

Source takeaway

Change one thing at a time and re-run /context in a fresh session. Turning off four plugins (claude-mem, claude-memory, Caveman, understand-anything) cut custom agents from 943 tokens to 74; removing one of the author's own skills saved only about 100.

Topic links

Related projects

Tools that match this angle

#1Direct
Routing

LiteLLM

BerriAI/litellm

An OpenAI-compatible gateway and SDK for calling many model providers with budgets, logging, load balancing, guardrails, and cost tracking.

59.1K11.6KSource-available
gatewaycost-trackingrouting
#2Direct
Observability

Langfuse

langfuse/langfuse

Open-source LLM engineering platform for observability, traces, metrics, evals, prompt management, datasets, and playground workflows.

34.8K3.8KSource-available
tracesevalscosts
#4In spirit
Agents

LangGraph

langchain-ai/langgraph

A framework for building resilient stateful agents with explicit graphs, persistence, human-in-the-loop flows, and controllable execution.

41.9K7.1KMIT
agentsstateworkflows
Related feed

More source-linked context

XDA source artwork
long-formX
long-form

Dropping Claude Code from High to Medium effort cut output tokens 45%

XDA's Mahnoor Faisal ran five coding jobs on Sonnet 5 twice from an identical starting codebase, changing only the effort level. High spent about 26,000 output tokens; Medium finished the same work on roughly 14,300.

coding-agentstoken-consumptionai-spend
Read note
theclimatebrink.com source artwork
newsT
news

The real energy use of agentic AI

Climate scientist Zeke Hausfather metered his own Claude Code habit: 1,138 typed prompts fanned out to more than 14,000 model calls and 3.2 billion tokens in eight weeks, drawing roughly 170 kWh of data-center electricity.

tokenmaxxingcoding-agentsagents
Read note
Ars Technica source artwork
newsAT
news

Anthropic "pauses" token-based billing for its Claude Agent SDK

Anthropic paused its plan to move Claude Agent SDK power users onto metered API pricing, updating its billing page to put the rollout on hold while it reworks how heavy agent usage is charged on subscription plans.

tokenmaxxingcoding-agentsagents
Read note