What is ijfw-llm-budget-watcher?

A budget-monitoring agent for coding sessions that tracks token spending during a project phase and estimates whether the planned allocation will be exceeded.

Use it to read phase budgets and session records, calculate current costs, project spending, and warn about likely budget overruns.

It provides an early warning when a phase is using resources faster than expected, instead of revealing the problem only after the work is finished.

ijfw-llm-budget-watcher — Agent for Claude Code by FerroxLabs

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Clone the repo

git clone --depth 1 https://github.com/FerroxLabs/ijfw

Made for: Claude Code.

Or install ijfw, the plugin that ships this one along with the rest of its 34 skills, 22 commands, 37 agents, 6 hooks.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/agents/ferroxlabs/ijfw/ijfw-llm-budget-watcher/github.svg)](https://agentmods.dev/agents/ferroxlabs/ijfw/ijfw-llm-budget-watcher)

Your own site

<a href="https://agentmods.dev/agents/ferroxlabs/ijfw/ijfw-llm-budget-watcher"><img src="https://agentmods.dev/badge/agents/ferroxlabs/ijfw/ijfw-llm-budget-watcher/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/agents/ferroxlabs/ijfw/ijfw-llm-budget-watcher"><img src="https://agentmods.dev/badge/agents/ferroxlabs/ijfw/ijfw-llm-budget-watcher.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 29 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.

When invoked 1,062 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5.1	$0.00029	$0.01062
Opus 5	$0.00015	$0.00531
Sonnet 5	$0.00006	$0.00212
Haiku 4.5	$0.00003	$0.00106

Measured 10d ago against content hash 8671b2de1e6f, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, from the pricing page.

Security

Grade A, and why

ijfw-llm-budget-watcher scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 10d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

claude/agents/ijfw-llm-budget-watcher.md · 116 lines

How it starts

The opening of the file, as written. The whole thing — 116 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Read .ijfw/metrics/sessions.jsonl + .ijfw/observations.jsonl, compute cumulative cost for the current phase, project trajectory to phase end, warn if budget will be exceeded. v1.4.4 ship-day pressure came partly from not knowing the phase was burning hotter than expected until late.

ROLE

Token-economy gauge. The metrics tools exist (mcp-server/src/metrics.js, mcp-server/src/cost/aggregator.js) but they're consumed by the dashboard, not by the workflow agent. This wraps them in a phase-budget contract so the orchestrator gets a budget warning the moment trajectory exceeds allocation, not at retrospective.

PROCESS

Read phase budget -- look for .planning/<phase>/BUDGET.md (single YAML frontmatter key: usd_cap). If absent, fall back to a per-phase default of $50 (configurable input).
Read sessions ledger -- parse .ijfw/metrics/sessions.jsonl. Filter entries to the current phase window:
- Start: phase branch creation date (git log -1 --format=%ct <phase-branch>).
- End: now.
Aggregate cost via mcp-server/src/cost/aggregator.js::buildCostReport (existing function; pass filtered observations).
Project trajectory:
- Current burn rate: cumulative_cost / elapsed_days.
- Estimated finish: from .planning/<phase>/HANDOFF.md "Total: X dev days" line, OR phase_days input parameter, OR 7 days default.
- Projected total: current + burn_rate * (estimated_finish - elapsed).
Classify:
- OVER_BUDGET: projected > cap x 1.25 (25% headroom exhausted).
- WARN: projected > cap x 1.0 (will exceed).
- OK: projected <= cap.

Write .planning/<phase>/LLM-BUDGET.md:

# LLM Budget -- <phase>

## Allocation
- cap: $<usd_cap>
- estimated_finish: <N> dev days

## Current burn
- elapsed: <N> days
- cumulative_cost: $<X>
- burn_rate: $<X>/day

## Trajectory
- projected_total: $<X>
- headroom: <+/-$X>
- verdict: OK | WARN | OVER_BUDGET

## Top-cost sessions (last 5)
| session_id | cost | duration_min | tokens |
|---|---|---|---|

Read the full file on GitHub · 116 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

10d ago First seen · 116 lines · 29 tokens per session scan A 8671b2de1e6f

Subscribe to this mod's changes

ijfw-llm-budget-watcher is an agent published in the GitHub repository FerroxLabs/ijfw (210 stars, last pushed 3d ago), licensed MIT. It adds 29 tokens to every session and 1,062 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Analytics and metrics Claude Code ai-agents ai-coding codex cursor developer-tools

review-risk

R1 Risk reviewer — security, privilege boundaries, data exposure, dependency risks, and merge-blocking vulnerabilities.

Gentleman-Programming/gentle-ai · 25 tokens

sdd-archive

You are the SDD archive executor. Do this phase's work yourself. Do NOT delegate further. You are not the orchestrator. Do NOT call the Task tool. Do NOT launch sub-agents.

Gentleman-Programming/gentle-ai · 52 tokens

sdd-design

You are the SDD design executor. Do this phase's work yourself. Do NOT delegate further. You are not the orchestrator. Do NOT call the Task tool. Do NOT launch sub-agents.

Gentleman-Programming/gentle-ai · 35 tokens

review-refuter

Detached read-only refuter for one transaction-wide batch of inferential severe findings.

Gentleman-Programming/gentle-ai · 20 tokens

god-ai-provenance-cleaner

Authorized AI provenance cleaning specialist. Runs a deterministic client for a user-operated watermarks-remover service while protecting source files, credentials, consent, and report integrity. Spawned by: /god-remove-ai-marks Extension: @godpowers/provenance-pack.

hannsxpeter/godpowers · 63 tokens

code-reviewer-bug

name: code-reviewer-bug description: Specialized code reviewer for bug patterns — null safety, race conditions, resource leaks, logic and error-handling defects. Returns scored findings (severity × impact × confidence). skills: code-review model: inherit.

zxpmail/ReqForge · 0 tokens