agenttelemetry-app 2.1.0__tar.gz → 2.3.0__tar.gz

{agenttelemetry_app-2.1.0 → agenttelemetry_app-2.3.0}/MANIFEST.in

@@ -14,41 +14,36 @@ include native_activity.py
 include local_servers.py
 include local_servers_windows.py
 include parser.py
+include antigravity.py
+include cursor_data.py
 include release_notes.py
 include native.py
 include accounts.py
 include session_names.py
-include mcp_access.py
-include mcp_backend.py
-include mcp_protocol.py
-include mcp_server.py
-include mcp_tools.py
-include telemetry_query.py
-include telemetry_insights.py
 include index.html
 include chart.umd.min.js
 include manifest.json
 include sw.js
 include docs/python-package.md
-include docs/mcp.md
 include docs/release-notes.md
 include docs/native-activity.md
+include docs/antigravity.md
 include static/app.css
 include static/apple-touch-icon.png
 include static/charts.js
 include static/core.js
 include static/icon-192.png
 include static/icon-512.png
-include static/insights.js
 include static/maintenance.js
 include static/release-notes.js
-include static/mcp.js
 include static/timeline.js
 include static/tool-icons/LICENSE.hermes.txt
+include static/tool-icons/LICENSE.antigravity.txt
 include static/tool-icons/LICENSE.openclaw.txt
 include static/tool-icons/LICENSE.simple-icons.txt
 include static/tool-icons/README.md
 include static/tool-icons/claude.svg
+include static/tool-icons/antigravity.svg
 include static/tool-icons/copilot.svg
 include static/tool-icons/cursor.svg
 include static/tool-icons/hermes.svg

{agenttelemetry_app-2.1.0/agenttelemetry_app.egg-info → agenttelemetry_app-2.3.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: agenttelemetry-app
-Version: 2.1.0
+Version: 2.3.0
 Summary: A private, local dashboard for AI coding agent usage
 License-Expression: MIT
 Project-URL: Homepage, https://github.com/uttamdeb/agent-telemetry
@@ -29,26 +29,39 @@ logs your tools already write to your own machine and serves a clean, interactiv
 dashboard — estimated spend, tokens, active time, cache efficiency, an **Anthropic vs OpenAI
 vs Google** provider comparison, an activity calendar, and breakdowns by **model, provider,
 day, hour, weekday, tool, IDE, project and session** — plus how much **disk** all these logs
-eat, and suggestions for spending less, drawn from your own numbers.
+eat.
 
 By default, log inventory, usage and the cache stay on this machine. Separate actions can
-share selected exports with [paired devices](#your-devices-optional), return requested
-analytics to an enabled [local MCP client](docs/mcp.md) (which may forward tool results to
-its model provider), contact a provider after a confirmed **Read once** account lookup, or
+share selected exports with [paired devices](#your-devices-optional), contact a provider
+after a confirmed **Read once** account lookup, or
 fetch GitHub update and release-note metadata after you click **Check for updates** or
 **Load notes**. Cached release notes can be shown offline after an upgrade. No account or API key is needed
 for the dashboard; it has no runtime dependencies beyond Python's standard library and a
 vendored copy of Chart.js.
 
-Covers **Claude Code · Claude Desktop · Codex · GitHub Copilot · Cursor · OpenCode · Hermes Agent · OpenClaw**.
+Covers **Claude Code · Claude Desktop · Codex · GitHub Copilot · Cursor · Antigravity · OpenCode · Hermes Agent · OpenClaw**.
+
+**Get AgentTelemetry:** [Desktop apps for macOS and Windows](#download-the-app) ·
+[VS Code Marketplace](https://marketplace.visualstudio.com/items?itemName=agenttelemetry-team.agenttelemetry) ·
+[Open VSX](https://open-vsx.org/extension/agenttelemetry-team/agenttelemetry) (Cursor and compatible editors) ·
+[Browser dashboard (Python or source)](#quick-start) ·
+[Product page](https://agenttelemetry-app.vercel.app/)
+
+**AgentTelemetry Pro (optional).** Everything above stays free, open source and local. [AgentTelemetry
+Pro](https://agenttelemetry-app.vercel.app/#pro) is a separate add-on with a sign-in: your dashboard in any browser or
+on your phone; chat with your running Claude Code and Codex sessions from the web, your phone or the Pro app on Mac
+and Windows; answering Claude Code's permission prompts from your phone; sessions that message each other across
+your devices; Optimize; and the local MCP bridge. See [Free vs Pro](https://agenttelemetry-app.vercel.app/#pro).
 
 ## Download the app
 
-| Platform | Download |
+Choose the installer for your platform from the release's **Assets** section.
+
+| Platform | Release downloads |
 | --- | --- |
-| macOS 13+ · Apple Silicon | **[Download DMG](https://github.com/uttamdeb/agent-telemetry/releases/download/v2.1.0/AgentTelemetry-2.1.0-macos-arm64.dmg)** |
-| macOS 13+ · Intel | **[Download DMG](https://github.com/uttamdeb/agent-telemetry/releases/download/v2.1.0/AgentTelemetry-2.1.0-macos-x86_64.dmg)** |
-| Windows 10/11 · x64 | **[Download EXE installer](https://github.com/uttamdeb/agent-telemetry/releases/download/v2.1.0/AgentTelemetry-2.1.0-windows-x64-setup.exe)** |
+| macOS 13+ · Apple Silicon | **[Find the Apple Silicon DMG](https://github.com/uttamdeb/agent-telemetry/releases/latest)** |
+| macOS 13+ · Intel | **[Find the Intel DMG](https://github.com/uttamdeb/agent-telemetry/releases/latest)** |
+| Windows 10/11 · x64 | **[Find the Windows EXE installer](https://github.com/uttamdeb/agent-telemetry/releases/latest)** |
 
 **Python is bundled — no terminal setup needed.** macOS: open the DMG, drag the app to
 Applications and open it. Windows: run the installer and open AgentTelemetry from Start.
@@ -61,10 +74,11 @@ starting monitoring. On macOS, opening the app manually also lets you import fir
 
 These builds are unsigned/unnotarized. macOS may require **System Settings → Privacy &
 Security → Open Anyway**; Windows SmartScreen may require **More info → Run anyway**.
+[Product and install page](https://agenttelemetry-app.vercel.app/) ·
 [All releases](https://github.com/uttamdeb/agent-telemetry/releases) ·
-[What's new in 2.1.0](docs/releases/v2.1.0.md) ·
-[SHA-256 checksums](https://github.com/uttamdeb/agent-telemetry/releases/download/v2.1.0/SHA256SUMS.txt) ·
-[PyPI package 2.1.0](https://pypi.org/project/agenttelemetry-app/2.1.0/)
+[What's new in 2.3.0](docs/releases/v2.3.0.md) ·
+[SHA-256 checksums in release assets](https://github.com/uttamdeb/agent-telemetry/releases/latest) ·
+[PyPI package 2.3.0](https://pypi.org/project/agenttelemetry-app/2.3.0/)
 
 ---
 
@@ -166,12 +180,14 @@ estimated from tokens or combined across devices. Session details show recorded
 creator/account IDs with their attribution limits. Account metadata is private to the
 source device. See [setup, research and limitations](docs/account-limits.md).
 
-## Local servers (macOS)
+## Local servers (macOS and Windows)
 
-The **Local servers** tab and menu bar can list local project listeners, pin them,
-and save an explicit launch command for Start and Restart. Process actions ask for
-confirmation. They require a ready macOS backend; Windows and Linux show the support
-limitation. Commands stay on this machine and are never included in device sharing.
+The **Local servers** dashboard tab can list local project listeners, pin them,
+and save an explicit launch command for Start and Restart. macOS can also show pinned
+controls in the menu bar; the Windows tray opens the dashboard controls. Process actions
+ask for confirmation. Lifecycle controls are supported on macOS and Windows; Linux does
+not yet support this feature. Windows asks for a working folder and uses ownership-checked
+process controls. Commands stay on this machine and are never included in device sharing.
 Running a saved command gives it your account's file and network access.
 See [behavior, privacy and limitations](docs/local-servers.md).
 
@@ -238,25 +254,15 @@ retain one serialized snapshot per display tick, in memory only; the ledger is u
 as all writes. `POST /api/shutdown` additionally requires the owning Windows client's
 per-process control token; a normal standalone dashboard cannot be stopped by that route.
 
-## Local MCP access (optional)
-
-AgentTelemetry provides eight read-only MCP tools for usage, comparisons, supported insights,
-session summaries and timelines, data health, the telemetry catalog and saved subscription
-limits. MCP access is off by default. Enable it in Settings and confirm each scope. The local
-bridge requires an already-running dashboard and returns data only when the configured MCP
-client requests it. A cloud-backed client may send tool results to its provider; review that
-client's data-handling settings before enabling access. Project/device names and usage from
-connected devices require their own opt-in scopes. The MCP bridge never refreshes accounts,
-returns prompt text, or edits client configuration. See [MCP tools, setup and privacy
-limits](docs/mcp.md).
-
-The MCP bridge ships with the **2.1.0** application and Python package. It is not a separate
-marketplace listing; install/update AgentTelemetry, then use the [MCP setup guide](docs/mcp.md).
-
 > `--rebuild` deletes the cache, and with it every session whose log has since been deleted
 > from disk (see [Storage](#storage--what-these-logs-cost-you-in-disk)). Copy
 > `.usage_cache.json` somewhere safe first if you've cleaned up old logs.
 
+Deleting a session inside a tool doesn't remove it from your totals either. If OpenCode,
+Hermes or OpenClaw delete a session from their database, the dashboard keeps the usage it
+already counted and marks that session as deleted. The same applies to turns you undo in
+OpenCode and to a Claude or Copilot log that gets shorter.
+
 **Requirements:** Python 3.8+ on **macOS, Linux or Windows**. On Windows run
 `python dashboard.py` (or `run.cmd`); on macOS/Linux `python3 dashboard.py` (or `./run.sh`).
 
@@ -280,21 +286,33 @@ instead of cloning, delete `.usage_cache.json` first — that file is your perso
 
 ## Data sources
 
+**AgentTelemetry Pulsar** is the engine that reads these sources. It finds each tool's
+local logs, picks the usage signal out of them and turns every tool's format into the one
+shared usage shape the dashboard shows. It runs on your machine, and its code lives in
+`parser.py`.
+
 | Tool | Where it reads | Tokens |
 |---|---|---|
 | **Claude Code** | `~/.claude/projects/**/*.jsonl` | exact (in/out/cache read+write, 5m/1h tiers) |
 | **Claude Desktop** (agent mode) | `Claude/local-agent-mode-sessions/**` under App Support / `%APPDATA%` / `~/.config` | exact |
-| **Codex** | `~/.codex/sessions/**`, `~/.codex/archived_sessions/**` | exact (in/cached/out/reasoning); subagents are identified and labelled |
+| **Codex** | `~/.codex/sessions/**`, `~/.codex/archived_sessions/**` (plain `*.jsonl` and the zstd-compressed `*.jsonl.zst` Codex swaps in for idle threads) | exact (in/cached/out/reasoning); subagents are identified and labelled. Compressed rollouts are read with Python 3.14+ or the `zstd` command; without either they are counted as unread (never dropped) and the sidebar says how many |
 | **GitHub Copilot** | VS Code-family `chatSessions/*.{json,jsonl}`; SDK/CLI `~/.copilot/session-state/*/events.jsonl`, enriched from VS Code `agentSessionData/*/session.db` | recorded request/turn tokens where available; SDK shutdown totals where safely dated. SDK gaps show a warning tooltip and unavailable/partial usage; older non-SDK chats retain text estimates. Premium multipliers are separate from estimated dollars. |
-| **Cursor** (native AI) | `Cursor/User/globalStorage/state.vscdb` under App Support / `%APPDATA%` / `~/.config` | partial — model, mode, timestamps, tool calls and AI-line stats are exact; tokens are on only ~2% of messages |
+| **Cursor** (native AI) | `Cursor/User/globalStorage/state.vscdb` under App Support / `%APPDATA%` / `~/.config` | recorded, partial or unavailable per session/model: explicit message input/output tokens, models, tools and AI-line stats where the local records expose them |
+| **Google Antigravity** | `~/.gemini/antigravity/conversations/*.db`, `~/.gemini/antigravity-ide/conversations/*.db`, `~/.gemini/antigravity-cli/conversations/*.db` | recorded SQLite usage with per-field availability; standalone legacy `.pb` files are unsupported |
 | **OpenCode** | `~/.local/share/opencode/opencode.db`, `%LOCALAPPDATA%\opencode\opencode.db`, `~/.opencode/opencode.db` (or `$OPENCODE_DATA_DIR`) | V1/V2 SQLite; exact (in/out/reasoning/cache), prompts and tools; cost is read from OpenCode's own per-message value |
 | **Hermes Agent** | `~/.hermes/state.db` (or `$HERMES_HOME`, `%LOCALAPPDATA%\hermes`) | exact (in/out/cache/reasoning, per model) |
 | **OpenClaw** | `~/.openclaw/agents/<agent>/agent/openclaw-agent.sqlite` (or `$OPENCLAW_STATE_DIR`; also the older `~/.clawdbot`, `~/.moltbot`) and the older per-session `sessions/*.jsonl` | exact (in/out/cache read/cache write/reasoning, per response); its own logged cost is used only for a model with no verified price. Large events are stored zstd-compressed: read with the `zstd` command or Python 3.14+ |
 
 A tool you don't use simply contributes nothing. **Attribution is by tool, not by model** —
-a Claude or GPT model used *inside* Copilot/Cursor/OpenCode/Hermes/OpenClaw counts under that tool, and the
+a Claude, Gemini, or GPT model used *inside* Copilot/Cursor/Antigravity/OpenCode/Hermes/OpenClaw counts under that tool, and the
 Models table lists each `model × tool` row separately.
 
+Antigravity's supported SQLite schema, duplicate-row handling, recorded checkpoint detail,
+and format limits are documented in [Google Antigravity support](docs/antigravity.md).
+
+Cursor's local conversation metadata is documented in [Cursor support](docs/cursor.md),
+including its model/token coverage states, workspace associations, subagents, and limits.
+
 Copilot's newer SDK-based agent sessions use a different store from older Copilot Chat.
 AgentTelemetry joins their event history to VS Code's saved per-turn usage using native
 session and event IDs. Custom endpoint models, including Cloudflare GLM 5.3, still count
@@ -374,26 +392,22 @@ anything is saved. Close the session in its original tool before confirming, and
 that tool if needed. Token, cost, prompt, tool-call and compaction records stay intact.
 Previous title metadata is backed up privately in
 `AgentTelemetryNative/session-title-backups/`, separately from the usage ledger.
-This supports local Claude Code transcripts, Codex through its installed CLI’s
+This supports local Claude Code transcripts, Codex through its
 [official name API](https://learn.chatgpt.com/docs/app-server), and OpenCode’s SQLite
-session stores. Other formats, remote sessions and deleted logs cannot be edited.
-Recognized local Claude subagent transcripts instead offer **Name in AgentTelemetry**.
-This saves a private display name on this device without editing Claude's transcript,
-native subagent name, or parent session. The preview states this scope explicitly.
+session stores. Codex's program is found wherever its apps install it (the ChatGPT app,
+the Codex editor extension, the Codex desktop app on Windows, or a Codex CLI), with no
+need for it on your PATH. Remote sessions and deleted logs cannot be renamed.
+Recognized local Claude subagent transcripts, and sessions from tools whose own title
+can't be edited safely (Copilot, Cursor, Claude Desktop, Antigravity, Hermes, OpenClaw),
+instead offer **Name in AgentTelemetry**. This saves a private display name on this
+device without editing the tool's files, Claude's transcript, native subagent name, or
+parent session. The preview states this scope explicitly.
 These names live in `AgentTelemetryNative/session-display-names.json`, outside the
 usage ledger, summaries, and device sharing. Enter the source title again to remove
 a display-name override. See [subagent name behavior](docs/subagent-sessions.md).
 Tool/provider icons are bundled locally, including the OpenAI mark for Codex; no remote
 logo requests occur. Their sources and licenses are in `static/tool-icons/README.md`.
 
-**Optimize** — suggestions derived from your own logs, ranked by what they'd save: sessions
-re-reading a very large context, thinking share, tool-heavy sessions that never delegated,
-cache written but never read, a costly model doing light work, what each **Skill** costs,
-edits that needed a retry, oversized **CLAUDE.md / AGENTS.md**, sessions that open heavier
-than your leanest ones, wasted file reads, installed skills and agents you never use, and
-MCP servers you never call, barely use, or load everywhere but use in one project. Nothing is shown unless
-your data supports it, and because the estimates overlap they are never summed.
-
 **Storage** — see below.
 
 **Keyboard** — `1`/`7`/`3`/`9`/`a` ranges, `m` month-to-date, `,` / `.` previous / next
@@ -475,26 +489,32 @@ dollars**, so cost is always derived. Rates live in `parser.py → PRICING` as
 `(input, output, cache_write_5m, cache_write_1h, cache_read)` per 1M tokens; edit freely
 (recomputed on each request, no re-parse needed).
 
-- **Anthropic** rates are current list prices (Opus 5.5 $4/$20, Opus 5 & 4.x $5/$25, Sonnet 5
-  & 5.5 $2/$10, Sonnet 4.x $3/$15, Haiku 4.5 $1/$5; cache write 1.25×/2× input for 5-min/1-hour, cache
-  read 0.1× — 0.05× on Opus 5.5). **OpenAI** GPT-5.4/5.5/5.6/6 are verified from OpenAI docs;
+- **Anthropic** rates are current list prices (Fable & Mythos 5.2 $8/$40, Fable & Mythos 5/5.1
+  $10/$50, Opus 5.5 $4/$20, Opus 5 & 4.x $5/$25, Sonnet 5 & 5.5 $2/$10, Sonnet 4.x $3/$15, Haiku
+  4.5 $1/$5; cache write 1.25×/2× input for 5-min/1-hour, cache read 0.1× — 0.05× on Opus 5.5 and
+  Sonnet 5.5, 0.025× on Fable & Mythos 5.1/5.2). **OpenAI** GPT-5.4/5.5/5.6/6 are verified from OpenAI docs;
   older/other models are estimates.
 - **Haiku 5.5** uses [Anthropic's published rates](https://platform.claude.com/docs/en/about-claude/pricing):
   $0.10/$0.50 per million input/output tokens up to 100K prompt tokens, $0.50/$2.50
   above 100K. The threshold includes cached input and applies to each request.
   Available Claude logs backfill this pricing detail on a normal restart; token and
   message counts stay unchanged. Older archives/peers and other tools without that
-  detail use the short-prompt rate. Optimize model substitutions conservatively use
-  the higher tier when the target request size is unknown.
+  detail use the short-prompt rate.
 - **Price changes are date-aware.** When a vendor cuts a price (OpenAI's GPT-5.6 cuts of
-  2026-07-30 and 2026-08-21), usage from before the change keeps the rate it had then —
+  2026-07-30 and 2026-08-21, Anthropic's Sonnet 5.5 cache-read cut of 2026-10-07), usage
+  from before the change keeps the rate it had then —
   see `parser.py → PRICE_HISTORY`.
 - **These are API-equivalent values.** If you're on a subscription (Claude Max/Pro, Codex,
   Copilot), you don't pay per token — the $ is "what this would cost at API rates."
-- **Cursor** rarely logs token counts, and **older Copilot chats** have none, so their tokens
-  (and thus $) are rough. Current Copilot builds log real per-request tokens, which are used.
-  Copilot's honest metric is **request count** and its **premium-request** total (both shown);
-  Cursor's is **messages, tool calls and AI lines kept** (also shown).
+- **Cursor** exposes explicit input/output token receipts for some messages; missing and
+  placeholder values are unavailable, and usage coverage can be partial. Cache and reasoning
+  values remain unavailable unless recorded. Overview shows observed Cursor prompts, replies,
+  and tool calls separately when they have no token/cost coverage. It does not estimate tokens
+  from text or context meters, infer subscription charges, or count metadata as tokens. The
+  **Exact tokens only** filter omits Cursor because local token coverage is incomplete. See
+  [Cursor support](docs/cursor.md). **Older Copilot chats** may have no token data; current
+  Copilot builds log real per-request tokens. Copilot's honest metric is **request count** and
+  its **premium-request** total (both shown); Cursor also exposes **AI lines kept**.
 - **Billing details that change the price are applied**: prompt-cache writes at their own
   rate (Claude and GPT-5.6/GPT-6), Claude fast mode, US-only inference (1.1x) and web
   searches ($10/1K). A forked or subagent Codex thread's replay of its parent is skipped,
@@ -504,12 +524,15 @@ dollars**, so cost is always derived. Rates live in `parser.py → PRICING` as
   when it runs through supported tools, at the Standard rate for prompts up to 200K
   tokens. Live, TTS, image and embedding models are not priced, and nor are the higher Pro rate
   for longer prompts or audio input. The Gemini CLI itself writes no token counts to read.
-- **Gemini 4 Argon** is recognized across the supported tools and uses Google's
-  [announced introductory rates](https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/):
-  $2 input, $10 output and $0.10 cached input per 1M tokens. Access is still limited to
-  trusted testers, and Google has not listed a public API model ID. These are API-equivalent
-  estimates for Argon usage found in logs. The later $4/$20 rate has no effective date yet
-  and is not applied ahead of time; cache storage charges are not modeled.
+- **Gemini 4 Argon and Gemini 4 Carbon** (`gemini-4-argon`, `gemini-4-carbon`) are recognized
+  across the supported tools and use Google's [published rates](https://aistudio.google.com/docs/pricing):
+  $2 input, $10 output and $0.10 cached input per 1M tokens through November 30, 2026. Google
+  doubles them to $4, $20 and $0.20 on December 1, 2026. The higher rate is not applied ahead
+  of time: it goes in once it takes effect, and earlier usage keeps the rate it had then.
+  Cache storage charges are not modeled.
+- **Gemini 4 Hydrogen** (`gemini-4-hydrogen`) uses Google's paid rate: $1.50 input, $7.50 output
+  and $0.075 cached input per 1M tokens. Its free tier can't be told apart in logs, so usage
+  is shown at the paid rate.
 
 ## Note on log retention
 
@@ -539,7 +562,7 @@ and extend this from that file).
 
 ## Files
 
-`dashboard.py` (server + cache + cost + `/api/storage`) · `parser.py` (log parsers + pricing) ·
+`dashboard.py` (server + cache + cost + `/api/storage`) · `parser.py` (AgentTelemetry Pulsar: log parsers + pricing) ·
 `index.html` (shell) · `static/app.css` · `static/core.js` · `static/charts.js` ·
 `static/views.js` · `chart.umd.min.js` (vendored Chart.js) · `AGENTS.md` · `run.sh`.
 
@@ -571,8 +594,8 @@ What sharing sends: session titles (often the start of a prompt), project and br
 log file paths, models, token counts, costs and times. Full prompts, replies and file
 contents are never sent. Sharing opens one read-only port (7879) that answers only to the
 pairing code. The dashboard itself stays on `127.0.0.1`. The connection is **not
-encrypted**, so share only on a network you trust. The setup checks on the Optimize tab
-(instruction files, installed skills, MCP servers) and the Storage tab still describe only
+encrypted**, so share only on a network you trust. The Storage tab and setup inventories
+(instruction files, installed skills, MCP servers) still describe only
 the computer you're looking at. `.peers.json` and `.peers/` are gitignored.
 
 **If it can't connect**, the error says which of these it is:

{agenttelemetry_app-2.1.0 → agenttelemetry_app-2.3.0}/README.md

@@ -5,26 +5,39 @@ logs your tools already write to your own machine and serves a clean, interactiv
 dashboard — estimated spend, tokens, active time, cache efficiency, an **Anthropic vs OpenAI
 vs Google** provider comparison, an activity calendar, and breakdowns by **model, provider,
 day, hour, weekday, tool, IDE, project and session** — plus how much **disk** all these logs
-eat, and suggestions for spending less, drawn from your own numbers.
+eat.
 
 By default, log inventory, usage and the cache stay on this machine. Separate actions can
-share selected exports with [paired devices](#your-devices-optional), return requested
-analytics to an enabled [local MCP client](docs/mcp.md) (which may forward tool results to
-its model provider), contact a provider after a confirmed **Read once** account lookup, or
+share selected exports with [paired devices](#your-devices-optional), contact a provider
+after a confirmed **Read once** account lookup, or
 fetch GitHub update and release-note metadata after you click **Check for updates** or
 **Load notes**. Cached release notes can be shown offline after an upgrade. No account or API key is needed
 for the dashboard; it has no runtime dependencies beyond Python's standard library and a
 vendored copy of Chart.js.
 
-Covers **Claude Code · Claude Desktop · Codex · GitHub Copilot · Cursor · OpenCode · Hermes Agent · OpenClaw**.
+Covers **Claude Code · Claude Desktop · Codex · GitHub Copilot · Cursor · Antigravity · OpenCode · Hermes Agent · OpenClaw**.
+
+**Get AgentTelemetry:** [Desktop apps for macOS and Windows](#download-the-app) ·
+[VS Code Marketplace](https://marketplace.visualstudio.com/items?itemName=agenttelemetry-team.agenttelemetry) ·
+[Open VSX](https://open-vsx.org/extension/agenttelemetry-team/agenttelemetry) (Cursor and compatible editors) ·
+[Browser dashboard (Python or source)](#quick-start) ·
+[Product page](https://agenttelemetry-app.vercel.app/)
+
+**AgentTelemetry Pro (optional).** Everything above stays free, open source and local. [AgentTelemetry
+Pro](https://agenttelemetry-app.vercel.app/#pro) is a separate add-on with a sign-in: your dashboard in any browser or
+on your phone; chat with your running Claude Code and Codex sessions from the web, your phone or the Pro app on Mac
+and Windows; answering Claude Code's permission prompts from your phone; sessions that message each other across
+your devices; Optimize; and the local MCP bridge. See [Free vs Pro](https://agenttelemetry-app.vercel.app/#pro).
 
 ## Download the app
 
-| Platform | Download |
+Choose the installer for your platform from the release's **Assets** section.
+
+| Platform | Release downloads |
 | --- | --- |
-| macOS 13+ · Apple Silicon | **[Download DMG](https://github.com/uttamdeb/agent-telemetry/releases/download/v2.1.0/AgentTelemetry-2.1.0-macos-arm64.dmg)** |
-| macOS 13+ · Intel | **[Download DMG](https://github.com/uttamdeb/agent-telemetry/releases/download/v2.1.0/AgentTelemetry-2.1.0-macos-x86_64.dmg)** |
-| Windows 10/11 · x64 | **[Download EXE installer](https://github.com/uttamdeb/agent-telemetry/releases/download/v2.1.0/AgentTelemetry-2.1.0-windows-x64-setup.exe)** |
+| macOS 13+ · Apple Silicon | **[Find the Apple Silicon DMG](https://github.com/uttamdeb/agent-telemetry/releases/latest)** |
+| macOS 13+ · Intel | **[Find the Intel DMG](https://github.com/uttamdeb/agent-telemetry/releases/latest)** |
+| Windows 10/11 · x64 | **[Find the Windows EXE installer](https://github.com/uttamdeb/agent-telemetry/releases/latest)** |
 
 **Python is bundled — no terminal setup needed.** macOS: open the DMG, drag the app to
 Applications and open it. Windows: run the installer and open AgentTelemetry from Start.
@@ -37,10 +50,11 @@ starting monitoring. On macOS, opening the app manually also lets you import fir
 
 These builds are unsigned/unnotarized. macOS may require **System Settings → Privacy &
 Security → Open Anyway**; Windows SmartScreen may require **More info → Run anyway**.
+[Product and install page](https://agenttelemetry-app.vercel.app/) ·
 [All releases](https://github.com/uttamdeb/agent-telemetry/releases) ·
-[What's new in 2.1.0](docs/releases/v2.1.0.md) ·
-[SHA-256 checksums](https://github.com/uttamdeb/agent-telemetry/releases/download/v2.1.0/SHA256SUMS.txt) ·
-[PyPI package 2.1.0](https://pypi.org/project/agenttelemetry-app/2.1.0/)
+[What's new in 2.3.0](docs/releases/v2.3.0.md) ·
+[SHA-256 checksums in release assets](https://github.com/uttamdeb/agent-telemetry/releases/latest) ·
+[PyPI package 2.3.0](https://pypi.org/project/agenttelemetry-app/2.3.0/)
 
 ---
 
@@ -142,12 +156,14 @@ estimated from tokens or combined across devices. Session details show recorded
 creator/account IDs with their attribution limits. Account metadata is private to the
 source device. See [setup, research and limitations](docs/account-limits.md).
 
-## Local servers (macOS)
+## Local servers (macOS and Windows)
 
-The **Local servers** tab and menu bar can list local project listeners, pin them,
-and save an explicit launch command for Start and Restart. Process actions ask for
-confirmation. They require a ready macOS backend; Windows and Linux show the support
-limitation. Commands stay on this machine and are never included in device sharing.
+The **Local servers** dashboard tab can list local project listeners, pin them,
+and save an explicit launch command for Start and Restart. macOS can also show pinned
+controls in the menu bar; the Windows tray opens the dashboard controls. Process actions
+ask for confirmation. Lifecycle controls are supported on macOS and Windows; Linux does
+not yet support this feature. Windows asks for a working folder and uses ownership-checked
+process controls. Commands stay on this machine and are never included in device sharing.
 Running a saved command gives it your account's file and network access.
 See [behavior, privacy and limitations](docs/local-servers.md).
 
@@ -214,25 +230,15 @@ retain one serialized snapshot per display tick, in memory only; the ledger is u
 as all writes. `POST /api/shutdown` additionally requires the owning Windows client's
 per-process control token; a normal standalone dashboard cannot be stopped by that route.
 
-## Local MCP access (optional)
-
-AgentTelemetry provides eight read-only MCP tools for usage, comparisons, supported insights,
-session summaries and timelines, data health, the telemetry catalog and saved subscription
-limits. MCP access is off by default. Enable it in Settings and confirm each scope. The local
-bridge requires an already-running dashboard and returns data only when the configured MCP
-client requests it. A cloud-backed client may send tool results to its provider; review that
-client's data-handling settings before enabling access. Project/device names and usage from
-connected devices require their own opt-in scopes. The MCP bridge never refreshes accounts,
-returns prompt text, or edits client configuration. See [MCP tools, setup and privacy
-limits](docs/mcp.md).
-
-The MCP bridge ships with the **2.1.0** application and Python package. It is not a separate
-marketplace listing; install/update AgentTelemetry, then use the [MCP setup guide](docs/mcp.md).
-
 > `--rebuild` deletes the cache, and with it every session whose log has since been deleted
 > from disk (see [Storage](#storage--what-these-logs-cost-you-in-disk)). Copy
 > `.usage_cache.json` somewhere safe first if you've cleaned up old logs.
 
+Deleting a session inside a tool doesn't remove it from your totals either. If OpenCode,
+Hermes or OpenClaw delete a session from their database, the dashboard keeps the usage it
+already counted and marks that session as deleted. The same applies to turns you undo in
+OpenCode and to a Claude or Copilot log that gets shorter.
+
 **Requirements:** Python 3.8+ on **macOS, Linux or Windows**. On Windows run
 `python dashboard.py` (or `run.cmd`); on macOS/Linux `python3 dashboard.py` (or `./run.sh`).
 
@@ -256,21 +262,33 @@ instead of cloning, delete `.usage_cache.json` first — that file is your perso
 
 ## Data sources
 
+**AgentTelemetry Pulsar** is the engine that reads these sources. It finds each tool's
+local logs, picks the usage signal out of them and turns every tool's format into the one
+shared usage shape the dashboard shows. It runs on your machine, and its code lives in
+`parser.py`.
+
 | Tool | Where it reads | Tokens |
 |---|---|---|
 | **Claude Code** | `~/.claude/projects/**/*.jsonl` | exact (in/out/cache read+write, 5m/1h tiers) |
 | **Claude Desktop** (agent mode) | `Claude/local-agent-mode-sessions/**` under App Support / `%APPDATA%` / `~/.config` | exact |
-| **Codex** | `~/.codex/sessions/**`, `~/.codex/archived_sessions/**` | exact (in/cached/out/reasoning); subagents are identified and labelled |
+| **Codex** | `~/.codex/sessions/**`, `~/.codex/archived_sessions/**` (plain `*.jsonl` and the zstd-compressed `*.jsonl.zst` Codex swaps in for idle threads) | exact (in/cached/out/reasoning); subagents are identified and labelled. Compressed rollouts are read with Python 3.14+ or the `zstd` command; without either they are counted as unread (never dropped) and the sidebar says how many |
 | **GitHub Copilot** | VS Code-family `chatSessions/*.{json,jsonl}`; SDK/CLI `~/.copilot/session-state/*/events.jsonl`, enriched from VS Code `agentSessionData/*/session.db` | recorded request/turn tokens where available; SDK shutdown totals where safely dated. SDK gaps show a warning tooltip and unavailable/partial usage; older non-SDK chats retain text estimates. Premium multipliers are separate from estimated dollars. |
-| **Cursor** (native AI) | `Cursor/User/globalStorage/state.vscdb` under App Support / `%APPDATA%` / `~/.config` | partial — model, mode, timestamps, tool calls and AI-line stats are exact; tokens are on only ~2% of messages |
+| **Cursor** (native AI) | `Cursor/User/globalStorage/state.vscdb` under App Support / `%APPDATA%` / `~/.config` | recorded, partial or unavailable per session/model: explicit message input/output tokens, models, tools and AI-line stats where the local records expose them |
+| **Google Antigravity** | `~/.gemini/antigravity/conversations/*.db`, `~/.gemini/antigravity-ide/conversations/*.db`, `~/.gemini/antigravity-cli/conversations/*.db` | recorded SQLite usage with per-field availability; standalone legacy `.pb` files are unsupported |
 | **OpenCode** | `~/.local/share/opencode/opencode.db`, `%LOCALAPPDATA%\opencode\opencode.db`, `~/.opencode/opencode.db` (or `$OPENCODE_DATA_DIR`) | V1/V2 SQLite; exact (in/out/reasoning/cache), prompts and tools; cost is read from OpenCode's own per-message value |
 | **Hermes Agent** | `~/.hermes/state.db` (or `$HERMES_HOME`, `%LOCALAPPDATA%\hermes`) | exact (in/out/cache/reasoning, per model) |
 | **OpenClaw** | `~/.openclaw/agents/<agent>/agent/openclaw-agent.sqlite` (or `$OPENCLAW_STATE_DIR`; also the older `~/.clawdbot`, `~/.moltbot`) and the older per-session `sessions/*.jsonl` | exact (in/out/cache read/cache write/reasoning, per response); its own logged cost is used only for a model with no verified price. Large events are stored zstd-compressed: read with the `zstd` command or Python 3.14+ |
 
 A tool you don't use simply contributes nothing. **Attribution is by tool, not by model** —
-a Claude or GPT model used *inside* Copilot/Cursor/OpenCode/Hermes/OpenClaw counts under that tool, and the
+a Claude, Gemini, or GPT model used *inside* Copilot/Cursor/Antigravity/OpenCode/Hermes/OpenClaw counts under that tool, and the
 Models table lists each `model × tool` row separately.
 
+Antigravity's supported SQLite schema, duplicate-row handling, recorded checkpoint detail,
+and format limits are documented in [Google Antigravity support](docs/antigravity.md).
+
+Cursor's local conversation metadata is documented in [Cursor support](docs/cursor.md),
+including its model/token coverage states, workspace associations, subagents, and limits.
+
 Copilot's newer SDK-based agent sessions use a different store from older Copilot Chat.
 AgentTelemetry joins their event history to VS Code's saved per-turn usage using native
 session and event IDs. Custom endpoint models, including Cloudflare GLM 5.3, still count
@@ -350,26 +368,22 @@ anything is saved. Close the session in its original tool before confirming, and
 that tool if needed. Token, cost, prompt, tool-call and compaction records stay intact.
 Previous title metadata is backed up privately in
 `AgentTelemetryNative/session-title-backups/`, separately from the usage ledger.
-This supports local Claude Code transcripts, Codex through its installed CLI’s
+This supports local Claude Code transcripts, Codex through its
 [official name API](https://learn.chatgpt.com/docs/app-server), and OpenCode’s SQLite
-session stores. Other formats, remote sessions and deleted logs cannot be edited.
-Recognized local Claude subagent transcripts instead offer **Name in AgentTelemetry**.
-This saves a private display name on this device without editing Claude's transcript,
-native subagent name, or parent session. The preview states this scope explicitly.
+session stores. Codex's program is found wherever its apps install it (the ChatGPT app,
+the Codex editor extension, the Codex desktop app on Windows, or a Codex CLI), with no
+need for it on your PATH. Remote sessions and deleted logs cannot be renamed.
+Recognized local Claude subagent transcripts, and sessions from tools whose own title
+can't be edited safely (Copilot, Cursor, Claude Desktop, Antigravity, Hermes, OpenClaw),
+instead offer **Name in AgentTelemetry**. This saves a private display name on this
+device without editing the tool's files, Claude's transcript, native subagent name, or
+parent session. The preview states this scope explicitly.
 These names live in `AgentTelemetryNative/session-display-names.json`, outside the
 usage ledger, summaries, and device sharing. Enter the source title again to remove
 a display-name override. See [subagent name behavior](docs/subagent-sessions.md).
 Tool/provider icons are bundled locally, including the OpenAI mark for Codex; no remote
 logo requests occur. Their sources and licenses are in `static/tool-icons/README.md`.
 
-**Optimize** — suggestions derived from your own logs, ranked by what they'd save: sessions
-re-reading a very large context, thinking share, tool-heavy sessions that never delegated,
-cache written but never read, a costly model doing light work, what each **Skill** costs,
-edits that needed a retry, oversized **CLAUDE.md / AGENTS.md**, sessions that open heavier
-than your leanest ones, wasted file reads, installed skills and agents you never use, and
-MCP servers you never call, barely use, or load everywhere but use in one project. Nothing is shown unless
-your data supports it, and because the estimates overlap they are never summed.
-
 **Storage** — see below.
 
 **Keyboard** — `1`/`7`/`3`/`9`/`a` ranges, `m` month-to-date, `,` / `.` previous / next
@@ -451,26 +465,32 @@ dollars**, so cost is always derived. Rates live in `parser.py → PRICING` as
 `(input, output, cache_write_5m, cache_write_1h, cache_read)` per 1M tokens; edit freely
 (recomputed on each request, no re-parse needed).
 
-- **Anthropic** rates are current list prices (Opus 5.5 $4/$20, Opus 5 & 4.x $5/$25, Sonnet 5
-  & 5.5 $2/$10, Sonnet 4.x $3/$15, Haiku 4.5 $1/$5; cache write 1.25×/2× input for 5-min/1-hour, cache
-  read 0.1× — 0.05× on Opus 5.5). **OpenAI** GPT-5.4/5.5/5.6/6 are verified from OpenAI docs;
+- **Anthropic** rates are current list prices (Fable & Mythos 5.2 $8/$40, Fable & Mythos 5/5.1
+  $10/$50, Opus 5.5 $4/$20, Opus 5 & 4.x $5/$25, Sonnet 5 & 5.5 $2/$10, Sonnet 4.x $3/$15, Haiku
+  4.5 $1/$5; cache write 1.25×/2× input for 5-min/1-hour, cache read 0.1× — 0.05× on Opus 5.5 and
+  Sonnet 5.5, 0.025× on Fable & Mythos 5.1/5.2). **OpenAI** GPT-5.4/5.5/5.6/6 are verified from OpenAI docs;
   older/other models are estimates.
 - **Haiku 5.5** uses [Anthropic's published rates](https://platform.claude.com/docs/en/about-claude/pricing):
   $0.10/$0.50 per million input/output tokens up to 100K prompt tokens, $0.50/$2.50
   above 100K. The threshold includes cached input and applies to each request.
   Available Claude logs backfill this pricing detail on a normal restart; token and
   message counts stay unchanged. Older archives/peers and other tools without that
-  detail use the short-prompt rate. Optimize model substitutions conservatively use
-  the higher tier when the target request size is unknown.
+  detail use the short-prompt rate.
 - **Price changes are date-aware.** When a vendor cuts a price (OpenAI's GPT-5.6 cuts of
-  2026-07-30 and 2026-08-21), usage from before the change keeps the rate it had then —
+  2026-07-30 and 2026-08-21, Anthropic's Sonnet 5.5 cache-read cut of 2026-10-07), usage
+  from before the change keeps the rate it had then —
   see `parser.py → PRICE_HISTORY`.
 - **These are API-equivalent values.** If you're on a subscription (Claude Max/Pro, Codex,
   Copilot), you don't pay per token — the $ is "what this would cost at API rates."
-- **Cursor** rarely logs token counts, and **older Copilot chats** have none, so their tokens
-  (and thus $) are rough. Current Copilot builds log real per-request tokens, which are used.
-  Copilot's honest metric is **request count** and its **premium-request** total (both shown);
-  Cursor's is **messages, tool calls and AI lines kept** (also shown).
+- **Cursor** exposes explicit input/output token receipts for some messages; missing and
+  placeholder values are unavailable, and usage coverage can be partial. Cache and reasoning
+  values remain unavailable unless recorded. Overview shows observed Cursor prompts, replies,
+  and tool calls separately when they have no token/cost coverage. It does not estimate tokens
+  from text or context meters, infer subscription charges, or count metadata as tokens. The
+  **Exact tokens only** filter omits Cursor because local token coverage is incomplete. See
+  [Cursor support](docs/cursor.md). **Older Copilot chats** may have no token data; current
+  Copilot builds log real per-request tokens. Copilot's honest metric is **request count** and
+  its **premium-request** total (both shown); Cursor also exposes **AI lines kept**.
 - **Billing details that change the price are applied**: prompt-cache writes at their own
   rate (Claude and GPT-5.6/GPT-6), Claude fast mode, US-only inference (1.1x) and web
   searches ($10/1K). A forked or subagent Codex thread's replay of its parent is skipped,
@@ -480,12 +500,15 @@ dollars**, so cost is always derived. Rates live in `parser.py → PRICING` as
   when it runs through supported tools, at the Standard rate for prompts up to 200K
   tokens. Live, TTS, image and embedding models are not priced, and nor are the higher Pro rate
   for longer prompts or audio input. The Gemini CLI itself writes no token counts to read.
-- **Gemini 4 Argon** is recognized across the supported tools and uses Google's
-  [announced introductory rates](https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/):
-  $2 input, $10 output and $0.10 cached input per 1M tokens. Access is still limited to
-  trusted testers, and Google has not listed a public API model ID. These are API-equivalent
-  estimates for Argon usage found in logs. The later $4/$20 rate has no effective date yet
-  and is not applied ahead of time; cache storage charges are not modeled.
+- **Gemini 4 Argon and Gemini 4 Carbon** (`gemini-4-argon`, `gemini-4-carbon`) are recognized
+  across the supported tools and use Google's [published rates](https://aistudio.google.com/docs/pricing):
+  $2 input, $10 output and $0.10 cached input per 1M tokens through November 30, 2026. Google
+  doubles them to $4, $20 and $0.20 on December 1, 2026. The higher rate is not applied ahead
+  of time: it goes in once it takes effect, and earlier usage keeps the rate it had then.
+  Cache storage charges are not modeled.
+- **Gemini 4 Hydrogen** (`gemini-4-hydrogen`) uses Google's paid rate: $1.50 input, $7.50 output
+  and $0.075 cached input per 1M tokens. Its free tier can't be told apart in logs, so usage
+  is shown at the paid rate.
 
 ## Note on log retention
 
@@ -515,7 +538,7 @@ and extend this from that file).
 
 ## Files
 
-`dashboard.py` (server + cache + cost + `/api/storage`) · `parser.py` (log parsers + pricing) ·
+`dashboard.py` (server + cache + cost + `/api/storage`) · `parser.py` (AgentTelemetry Pulsar: log parsers + pricing) ·
 `index.html` (shell) · `static/app.css` · `static/core.js` · `static/charts.js` ·
 `static/views.js` · `chart.umd.min.js` (vendored Chart.js) · `AGENTS.md` · `run.sh`.
 
@@ -547,8 +570,8 @@ What sharing sends: session titles (often the start of a prompt), project and br
 log file paths, models, token counts, costs and times. Full prompts, replies and file
 contents are never sent. Sharing opens one read-only port (7879) that answers only to the
 pairing code. The dashboard itself stays on `127.0.0.1`. The connection is **not
-encrypted**, so share only on a network you trust. The setup checks on the Optimize tab
-(instruction files, installed skills, MCP servers) and the Storage tab still describe only
+encrypted**, so share only on a network you trust. The Storage tab and setup inventories
+(instruction files, installed skills, MCP servers) still describe only
 the computer you're looking at. `.peers.json` and `.peers/` are gitignored.
 
 **If it can't connect**, the error says which of these it is: