AgentNotch is live · free · last verified 24 Oct 2026 · viability monitor
AgentNotch
AgentNotch streams live Claude Code and Codex token and cost telemetry into your Mac's notch
A deliberately small tool that nails its one job. If you already run Claude Code or Codex on a notched MacBook, a brew cask plus one env var buys you live tool calls and rough spend without a window switch. Skip it if you code on Windows or Linux, drive Cursor or Gemini CLI, or need shared dashboards — it covers none of those.
Verdict generatedVerified 6d ago · liveness 68/100 · cite: rightaichoice.com/tools/agentnotch
- Developers running Claude Code or Codex daily who want live token and cost visibility
- MacBook Pro 14"/16" (2021+) owners who want glanceable agent status without leaving the editor
- Solo devs who prefer a native macOS app over tailing terminal logs
- Open-source fans comfortable setting an OTEL endpoint by hand
- Developers on Windows or Linux — macOS 14.0 (Sonoma) or later is required
- Anyone using Cursor, Aider, Gemini CLI or other assistants — only Claude Code and Codex report in
- Teams needing multi-user dashboards, SSO or SLAs — this is a single-machine local tool
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip AgentNotch if you're on Windows or Linux, use an assistant other than Claude Code or OpenAI Codex, or need trace-level spans and multi-user dashboards rather than a single-machine notch display.
The app itself is free and MIT-licensed, but you still pay your model provider for every token Claude Code or Codex consumes — AgentNotch only measures that spend, it doesn't reduce it.
AgentNotch is $0 with no feature gating, so there's no tier ladder to climb and no seat math. Compared with paid agent-observability platforms that charge per seat or per ingested event, it's the cheapest possible entry point for one developer on one Mac — but it also offers none of what those paid platforms sell: multi-user aggregation, hosted retention, or alerting.
In short
AgentNotch — AgentNotch streams live Claude Code and Codex token and cost telemetry into your Mac's notch. Best for Developers running Claude Code or Codex daily who want live token and cost visibility, MacBook Pro 14"/16" (2021+) owners who want glanceable agent status without leaving the editor, Solo devs who prefer a native macOS app over tailing terminal logs. Free to use.
What is AgentNotch?
AgentNotch — AgentNotch streams live Claude Code and Codex token and cost telemetry into your Mac's notch. Best for Developers running Claude Code or Codex daily who want live token and cost visibility, MacBook Pro 14"/16" (2021+) owners who want glanceable agent status without leaving the editor, Solo devs who prefer a native macOS app over tailing terminal logs. Free to use.
AgentNotch is a free, MIT-licensed macOS app that shows what your AI coding assistant is doing right now — from the notch of your MacBook. It listens for OTLP/HTTP telemetry on port 4318, decodes OTLP logs (/v1/logs) and metrics (/v1/metrics), and renders tool calls, input/output token counts and estimated spend as they arrive. If your Mac has no notch, it falls back to the menu bar. The interaction model is glance-first. Hover the notch to expand a short list of recent tool calls; click through for full history, a token breakdown and settings behind the gear icon. A completion notification tells you when a task wraps up, so you stop alt-tabbing to the terminal to check. Source-aware coloring keeps two assistants legible at once: orange for Claude Code, blue for Codex, light blue for anything unrecognized. Setup is one environment variable for Claude Code (export OTEL_EXPORTER_OTLP_ENDPOINT to http://localhost:4318) and an [otel] section in ~/.codex/config.toml for Codex CLI v0.79+, where the README suggests setting trace_exporter to "none" since the app doesn't decode traces. Install via Homebrew cask from the AppGram tap, or grab the release zip manually. Requirements are macOS 14.0 Sonoma or later. Everything is local — the app only receives telemetry and sends nothing out. Treat it as a single-machine activity and cost guardrail rather than an observability platform. Coverage stops at Claude Code and Codex, there are no trace-level analytics and no team dashboards, and cost figures are token-derived estimates. For span-level debugging or org-wide AI spend reporting, you'll need a hosted LLM observability product instead.
Key features
- Real-time tool call tracking for file reads, code edits and shell commands
- Input/output token monitoring with live cost estimates
- Completion notification when an AI task finishes
- Color-coded sources: orange Claude Code, blue Codex, light blue unknown
- Notch display on notched Macs with menu bar fallback
- Hover to expand recent tool calls with details
- Click to open full tool call history with token breakdown
- Toggle show/hide for token counts
- Toggle show/hide for cost estimates
- Filter telemetry by source (Claude Code / Codex)
- Toggle menu bar icon visibility
- OTLP/HTTP telemetry ingestion on port 4318
- Decodes OTLP logs (/v1/logs) and metrics (/v1/metrics)
- Native Swift macOS app with no Electron overhead
- Runs 100% locally — no telemetry leaves the machine
Behind the verdict
Where AgentNotch earns its place is the ambient check. Long agent runs create a specific anxiety — is it still working, and how much is this run costing me — and the notch is a better home for that answer than a terminal tab you keep flipping to. We'd reach for it when you're running Claude Code or Codex for hours at a stretch and want tool calls plus estimated spend in peripheral vision. Pass if your setup doesn't match the shape of the tool. macOS 14.0 or later is required, and non-notch Macs get the menu bar fallback rather than the notch itself. Only Claude Code and Codex report in — Cursor, Aider and Gemini CLI users get nothing to look at. The closest alternative is the terminal itself: tailing logs or reading your provider's usage dashboard. AgentNotch is nicer to glance at and cheaper than a hosted LLM observability platform, but it will not replace one. There's no trace or span analytics, no team view, no SSO, no SLA. It's one developer on one machine. Caveat on the numbers. Cost is derived from token counts, so it's an estimate, not your invoice. Useful for catching a runaway loop; not something to reconcile a budget against. Traces aren't decoded either, which is why the README tells you to disable trace export for Codex. Setup is the friction point, and it's modest: an env var for Claude Code, an [otel] block in ~/.codex/config.toml for Codex CLI v0.79+. In practice the install is a brew tap and cask, and the whole thing stays local — it receives telemetry and sends nothing out. If a hand-edited config file makes you flinch, that's the honest reason to skip it rather than anything about the app's quality.
Models under the hood
as of 2026-10-24
Limitations
- macOS-only menu bar app that lives in the Mac's notch (with a menu bar fallback on non-notch Macs).
- It surfaces telemetry specifically for Claude Code and OpenAI Codex.
- It's an early-stage open-source project (repository shows ~220 stars, 4 open issues) with limited external documentation.
as of 2026-10-04
Researching AgentNotch? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Is AgentNotch still active?
AgentNotch is live · free · last verified 24 Oct 2026 · viability monitor.
Viability score
How well maintained and how widely used is AgentNotch? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →What do users say about AgentNotch?
What people actually say about AgentNotch — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
ResearchedDiscussed by users across 2 sources (YouTube, Product Hunt) · researched Oct 20, 2026.
Weighted by the 15 posts each of 2 sources contributed.
- +Free and MIT-licensed, making it risk-free to try and modify
- +Native Swift with no Electron overhead, so it stays lightweight in the menu bar
- +Runs 100% locally; no telemetry leaves the machine, easing privacy concerns
- +Notch placement is genuinely novel and removes a habit of checking the terminal
- +Live token and cost estimates directly target API-bill anxiety for solo developers
- −Only covers Claude Code and Codex; other AI coding assistants are invisible
- −No trace decoding, so detailed request timelines are out of scope
- −Requires environment-variable and config-file setup that assumes some comfort with terminals
- −Unclear whether custom Codex tools are tracked, a question left unanswered in the data
- −No Xcode integration despite Mac developer-heavy target audience
- • No direct app costs, but the cost visibility it surfaces makes API spend harder to ignore
- • Time cost of setting environment variables and Codex config.toml
How we verified this
Every listed fact about AgentNotch— pricing, features, links, status — is re-read from the vendor's own pages by our verification pipeline and each change is logged with a date; the last pass ran on 24 Oct 2026. Pages mix model-assisted synthesis with human review; see how we verify.
Verification history
We have re-verified AgentNotch 11 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — changed pricing page, limitations, our verdict, description, features, pricing tiers, summary
- — changed our verdict, who should skip it, description, who it suits, features, summary, pricing tiers
- — changed pricing page, limitations
- — changed who it suits, description, features, pricing tiers, who should skip it, our verdict
- — changed pricing page
- — changed use cases, description, our verdict, features, hidden costs, limitations, switching in, switching out, pricing tiers, pricing fit, setup time, who should skip it, summary
Showing the 6 most recent of 11 verification passes.
Free to cite with attribution — this page re-verifies continuously.
How much does AgentNotch cost in 2026?
AgentNotch: Free to use. Pricing last verified 24 Oct 2026.
Plans compared
Every published AgentNotch tier, its price as the vendor publishes it, who it fits, and what it adds over the tier below. The 12-month cost projection follows.
| Plan | Price | Billing | Who it's for |
|---|---|---|---|
| Free | Free | — | Solo developers on a MacBook Pro 14"/16" (2021+) who run Claude Code or OpenAI Codex daily and want live token and cost visibility without adding a subscription.Adds: Starting tier — the whole app is free with no feature gating: real-time tool call, token and cost tracking, notch display, source filtering and completion notifications. |
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Where the pricing makes sense
The company stage and team size where AgentNotch's pricing actually pencils out — and where peers do it cheaper.
AgentNotch is $0 with no feature gating, so there's no tier ladder to climb and no seat math. Compared with paid agent-observability platforms that charge per seat or per ingested event, it's the cheapest possible entry point for one developer on one Mac — but it also offers none of what those paid platforms sell: multi-user aggregation, hosted retention, or alerting.
Has AgentNotch's pricing changed?
No plan price for AgentNotch has moved since we began logging in Aug 2026; the pricing page is re-read on every verification pass.
Who should pick AgentNotch?
Pick AgentNotch if you are Developers running Claude Code or Codex daily who want live token and cost visibility, MacBook Pro 14"/16" (2021+) owners who want glanceable agent status without leaving the editor, Solo devs who prefer a native macOS app over tailing terminal logs, Open-source fans comfortable setting an OTEL endpoint by hand.
Look elsewhere if you are Developers on Windows or Linux — macOS 14.0 (Sonoma) or later is required, Anyone using Cursor, Aider, Gemini CLI or other assistants — only Claude Code and Codex report in, Teams needing multi-user dashboards, SSO or SLAs — this is a single-machine local tool, Users who need trace- or span-level analytics — the app decodes logs and metrics, not traces.
Skip it if Skip AgentNotch if you're on Windows or Linux, use an assistant other than Claude Code or OpenAI Codex, or need trace-level spans and multi-user dashboards rather than a single-machine notch display.
GeneratedReal-world workflow fit
Concrete scenarios for the personas AgentNotch actually fits — and what changes day-one when you adopt it.
You export OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4318 in your Claude Code config, launch AgentNotch, and start a long refactor. You keep working in your editor while the notch shows an orange indicator and a running stack of file reads and edits; hovering shows the last few tool calls, and you leave token and cost toggles on so the spend is always visible.
Outcome: You notice the agent has been looping on the same file and cut the run early, and you finish the session knowing exactly what the run cost instead of discovering it on the next API bill.
You add [analytics] enabled = true and an [otel] section to ~/.codex/config.toml, point the otlp-http exporter at http://localhost:4318/v1/logs with protocol = "binary", and set trace_exporter = "none". Now AgentNotch shows a blue Codex indicator next to the orange Claude Code one, and you filter by source to compare what each assistant actually does during the same task.
Outcome: You get a per-source token and cost read on the same machine, and you can tell which assistant is cheaper for the kind of work you hand it.
You keep your collector and Grafana dashboards for historical and team-wide analysis, and add AgentNotch as the local front layer — each developer points their Claude Code and Codex OTLP endpoints at port 4318 on their own Mac so the notch lights up while they work.
Outcome: Individual developers get instant, zero-config visibility into the agent running on their machine, while the team keeps aggregate retention and dashboards in the existing stack.
Use cases
- Monitor Claude Code tool calls in real time from the macOS notch
- Watch OpenAI Codex as it reads files and executes commands
- Keep an eye on agent activity while working in other apps
- Debug AI coding assistant behavior by observing tool sequences
- Get passive visibility into agent progress without window switching
- Catch runaway token spend during a long agent run with live cost estimates
Setup time & first value
How long it actually takes to get something useful out of AgentNotch — broken out by persona, not the marketing-page minute.
Solo developer on a Mac: about 5 minutes — brew tap AppGram/tap, brew install --cask agentnotch, then set OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4318 for Claude Code. Codex users add a few lines to ~/.codex/config.toml (the [otel] block plus trace_exporter = "none"), which is another 5 minutes if you're comfortable editing TOML. First-value moment is the notch lighting up on your next agent
Switching to or from AgentNotch
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From terminal log monitoring: install the Homebrew cask, point OTEL_EXPORTER_OTLP_ENDPOINT at http://localhost:4318, and let the notch replace the scrolling output you were tailing.
- →From a self-hosted OpenTelemetry collector: keep the collector for retention and add AgentNotch as a local listener — it accepts OTLP/HTTP on port 4318 and needs no change to your existing export config beyond the
- →From nothing at all: brew tap AppGram/tap; brew install --cask agentnotch, then add the one-line OTEL endpoint export to your Claude Code configuration.
- ↗To Langfuse or Grafana: point your Claude Code and Codex OTLP exporters at the hosted or self-hosted collector instead of localhost:4318 — AgentNotch holds no data you need to export.
- ↗To a full observability platform: since AgentNotch stores nothing remotely and adds no instrumented code, migration is just repointing the OTLP endpoint; expect to give up the notch glanceability and cost estimates
Integrations
Resources & guides
Tutorials & Learning
YouTube returned 6 videos for “AgentNotch”, and we withheld 6: 6 could not be judged, because “AgentNotch” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about AgentNotch.
Official links
What are the best AgentNotch alternatives?
The closest alternatives to AgentNotch we list are Langfuse, Conan and Tokentelemetry. Ranked by direct product-type match, not category overlap.
| Alternative | Pricing | Starts at | Last verified |
|---|---|---|---|
| Langfuse | Freemium | Pricing not published | — |
| Conan | Freemium | Pricing not published | — |
| Tokentelemetry | Free | Free | — |
Alternatives to AgentNotch
View all
Langfuse
Open-source LLM observability: trace agents, manage prompts, and run evals on one MIT-licensed platform.
Conan
Conan is a native macOS HUD for Claude Code that streams every prompt, tool call, skill and token live as the session runs.
Tokentelemetry
Local, MIT-licensed token usage and cost tracking for AI coding agents — reads the logs your agents already write.
Tools that pair well with AgentNotch
Common stack mates teams adopt alongside AgentNotch, with the specific reason each pairing earns its keep.
Langfuse
Open-source LLM observability: trace agents, manage prompts, and run evals on one MIT-licensed platform.
Conan
Conan is a native macOS HUD for Claude Code that streams every prompt, tool call, skill and token live as the session runs.
Tokentelemetry
Local, MIT-licensed token usage and cost tracking for AI coding agents — reads the logs your agents already write.
AgentNotch vs its closest rival
AgentNotch's closest head-to-head is with Spider Cloud — Choose Spider Cloud if you need to supply your AI agents with fresh web data at scale — it's cheap, fast, and packed with AI-powered scraping features. Choose AgentNotch if you're a macOS developer who wants to keep an eye on Claude Code or Codex without leaving the notch. They solve totally different problems, so pick based on whether you need to gather data or monitor an assistant. Verdict generated
Head-to-head comparisons
Agentnotch vs Spider Cloud
Choose Spider Cloud if you need to supply your AI agents with fresh web data at scale — it's cheap, fast, and packed with AI-powered scraping features. Choose AgentNotch if you're a macOS developer who wants to keep an eye on Claude Code or Codex without leaving the notch. They solve totally different problems, so pick based on whether you need to gather data or monitor an assistant.
GeneratedAgentnotch vs Temporal Ai
Temporal AI and AgentNotch serve entirely different purposes. Temporal is an enterprise-grade durable execution platform for building reliable AI agents and complex workflows, ideal for teams needing fault tolerance and state persistence. AgentNotch is a niche macOS utility for passively monitoring Claude Code/Codex usage from the notch. Choose Temporal for production orchestration; choose AgentNotch for lightweight, local cost tracking of coding assistants. They are not direct competitors.
GeneratedAgentnotch vs Voyage Ai
Voyage AI and AgentNotch serve entirely different needs: Voyage AI delivers enterprise-grade embedding models for RAG pipelines, while AgentNotch provides a free, open-source macOS app for monitoring AI coding assistants. Choose Voyage if you're building a production RAG system on domain-specific data; choose AgentNotch if you're a macOS developer wanting real-time telemetry from Claude Code or Codex.
GeneratedFrequently asked questions
Categories
Used AgentNotch? Help shape our editorial sentiment research.