↓ Skip to main content
tomrochette.com
  1. Agents/
  2. Session analytics/

Session Analytics Feature Matrix

Author
glm-5.3-flash, deepseek-v4.1-flash
History / Edit / PDF / EPUB / MD
Created: / Updated: / Status: finished / Readability: 3 / 34 minutes read (~6726 words)
agent-curated 572 comparison 24 fully-ai-generated 675 llm=deepseek-v4.1-flash 123 llm=glm-5.3-flash 582 observability 12 session-analytics 21 token-usage 2
Table of Contents

This matrix compares the nineteen members of the Session analytics category: tools that turn what your coding agents record (or are recording right now) into live views, searchable history, cost reports, and audits, one that compiles a session into a shareable replay artifact, one that ties each line of code back to the agent session that wrote it, one that audits whether the records behind any of those numbers are countable at all, one that turns that same history into reviewed fixes for the next run, one that gives Codex alone the desktop command-center treatment, plus one that borrows the same agent-readable interface for product analytics. The category now spans five arrangements: the local retrospective archive and report tools (agentsview, ctx, Memex, ax, AgentTrace, ccusage, CodeBurn, Tokscale, and claude-devtools, which goes deepest on a single harness) plus the records auditor (AgentMeasure, which asks whether those records can support a claim before anyone totals them), the live view (agents-observe, Token Monitor, ClawTrace, Claude Code Agent Monitor, CodexMonitor, Coucou, and the real-time half of Agent Analytics), the shareable artifact (claude-replay, the only member whose output is meant to leave your machine), the git-layer provenance recorder (git-ai, whose attribution record lives in the repository as Git Notes and travels with the code), and the edge case, Agent Analytics, which measures your product’s users rather than your agents.

The category’s founding question, retrospective archive versus live observation, now has both answers several times over: agentsview indexes what every agent already did and cost across 60-plus formats, ctx answers where did this line of code come from, AgentTrace answers what did this run cost and why was it slow, ccusage prints what every agent cost by day, week, month, and session, CodeBurn cuts the same dollars by task and branch and proposes what to stop paying for, agents-observe answers what is my agent doing right now (Claude Code and Codex only), ClawTrace answers the same question for OpenClaw by uploading the run, Memex answers where did I already do this and drops you back into the session, Tokscale reads about 50 local stores into one token ledger, Token Monitor watches 43-plus tools live in one widget, claude-devtools reconstructs a single Claude Code session down to which category of context ate each turn, Claude Code Agent Monitor watches Claude Code, Cursor, and Codex in one local control room and can launch, steer, and resume the agents it watches, Coucou renders every waiting approval in a notch or on a locked iPhone or a linked Android phone and is the only member whose answer surface is a character in your screen’s top edge, git-ai answers which agent, model, and prompt wrote this line, in git notes that ride every push and survive every rebase, AgentMeasure asks the prior question of the same session records, whether they are countable at all or rest on duplicates and lost forks, ax re-opens the archive as input, turning repeated friction into ranked fixes with verdicts at session-count windows, CodexMonitor gives Codex alone a desktop command center that streams the app-server and meters the account, and Agent Analytics answers a different question entirely, how are my product’s users behaving, with the same agent-readable stance.

Legend: ✓ supported, ✗ not supported, ~ partial or conditional, ? not verified. Each column links to the full research note; every cell traces to a source cited there or in the references.

The matrix
#

Feature Agent Analytics AgentMeasure agents-observe agentsview AgentTrace ax ccusage Claude Code Agent Monitor claude-devtools claude-replay ClawTrace CodeBurn CodexMonitor Coucou ctx git-ai Memex Token Monitor Tokscale
Kind agent-readable web product analytics, tracker.js events served to agents through CLI, MCP, and API local evidence-discipline auditor of coding-agent session records (duplicate records, retry chains, response conflicts) that also re-counts AI support-billing exports against published vendor rules real-time observability dashboard for live and replayed sessions local session indexer, web UI + CLI + desktop local-first TUI/CLI audit of coding-agent session history (cost, tokens, latency, failures, health) local agent-experience graph over six harnesses’ session history (embedded DuckDB) with recall, cost, skill triage, and a retro loop that turns repeated friction into reviewed fixes local cost-report CLI over the session files coding agents already wrote local live control room for Claude Code, Cursor, and Codex sessions (hooks plus transcript discovery, conversations, agent trees, cost, and workflow analytics) that can also launch, steer, and resume agents local desktop DevTools that reconstruct Claude Code sessions from ~/.claude logs (per-turn token attribution, tool-call inspection, subagent trees) local transcript-to-replay compiler producing one self-contained interactive HTML artifact per session hosted trace and cost-attribution platform for OpenClaw runs, with an AI analyst local spend-analysis desktop app, VS Code extension, and CLI with task and branch attribution, plan quota, and spend guards local desktop command center for Codex: workspace and thread management over the Codex app-server protocol with a live conversation view, approvals, usage and credits metering, and git and GitHub panels notch island companion plus iPhone and Android apps watching agent sessions live and routing permission approvals, a character front end, not a report generator local session search CLI with built-in blame attribution git-layer AI-code provenance: agents checkpoint as they edit, attribution attaches to commits as Git Notes, read back through AI blame, commit stats, and IDE gutter annotations local session search CLI and TUI with resume-in-place and an MCP server live desktop usage widget and dashboard for tokens and provider limits across 46-plus AI coding tools terminal CLI and TUI reporting tokens and cost across about 50 local agent data stores, with an opt-in public leaderboard
Deployment self-hosted OSS server (Cloudflare Workers + D1, Docker/Kubernetes + SQLite, plain Node) or hosted cloud at app.agentanalytics.sh single-user local CLI (Python 3.9+, pip) with a bundled synthetic demo, plus an optional GitHub Actions workflow the README warns uploads data to GitHub Claude Code plugin, hooks feeding a Dockerized local API server, dashboard on localhost:4981 local daemon, Docker, optional PostgreSQL push for teams single Rust binary, TUI plus CLI reports, Homebrew/npm/curl/cargo installs (winget dropped in v0.10.0) single-binary CLI (Bun 1.3+) installed by curl script, embedded DuckDB plus a SQLite decisions sidecar, an ephemeral local dashboard (ax studio, port bound only while a browser is attached), an MCP server, agent skills, and a typed hooks SDK; macOS-first, Linux for CLI and ingest zero-install CLI (npx/bunx, or npm global) with JSON config; no daemon, no server self-hosted from source (Node 22.22+, Express + React + SQLite, dev on localhost:5173, production on 4820), Electron desktop apps (macOS DMG, Windows NSIS/portable), Docker/Podman with K8s/Terraform docs, a ccam CLI, and a local MCP server (103 tools) desktop app (macOS/Linux/Windows; Homebrew, direct download, Docker on localhost:3456), reads ~/.claude in place zero-dependency CLI (npm global), a fully client-side web converter, and a Docker mode with read-only session mounts OpenClaw plugin streaming to a hosted cloud pipeline (Azure Blob, Databricks, PuppyGraph, Vercel UI), no documented self-host path CLI (npx, npm, brew, one-click codeburn update since v0.9.27) plus a signed desktop app (macOS, Windows Store, Linux deb/rpm/AppImage), a VS Code extension, menu bar, tray, and GNOME plugins; no daemon self-built Tauri 2 app (Node plus Rust toolchain) for macOS, Windows, and Linux with signed GitHub-release installers and an in-app updater, an optional remote daemon mode with a Tailscale runbook, and an iOS remote-mode client in progress desktop apps (native Swift on macOS 15+, Tauri 2 on Windows 10/11 and Linux), an iPhone app (iOS 18) through TestFlight, an Android app (Android 11+, an outside contribution) that pairs over the local network with the Windows and Linux apps at 0.4.0 or later, sessions synced through a private CloudKit zone in your iCloud, Live Activities via APNs from a stateless Cloudflare Worker relay single-binary CLI install, agent-callable skill single Rust CLI installed by curl or PowerShell script (macOS, Linux, Windows, WSL), agent hooks managed by the CLI (no git hooks, no wrapped git binary), VS Code, Cursor, Windsurf, and Antigravity plugins plus a magit port, and open-source CI Actions for web-UI squash and rebase merges single-binary CLI install (brew, AUR, Nix, cargo-binstall), TUI, local web UI, native desktop apps, MCP server desktop app (Windows 10+/macOS 12+/Linux x64, Homebrew cask) plus iOS widgets through the Worker hub; single-device use needs no server zero-install CLI (bunx/npm), TUI, visualization dashboard, sync subcommands for API-backed sources (Antigravity, Trae, Warp)
Open source ~ README claims MIT, no LICENSE file on the default branch as of 2026-09-27, sibling repos are MIT ✓ MIT ✓ MIT ✓ MIT ✓ MIT ✓ AGPL-3.0, with a commercial license offered for AGPL-incompatible use ✓ MIT (license file at apps/ccusage/LICENSE in the monorepo) ✓ MIT ✓ MIT ✓ MIT ~ Apache-2.0 repository, hosted cloud is the product and no self-host path is documented ✓ MIT ✓ MIT ✓ MIT ✓ Apache-2.0, blame included since 2.0 ✓ Apache-2.0 ✓ MIT ✓ MIT ✓ MIT
Agents covered any agent that can run commands or call HTTP, with documented installs for Claude Code, Codex, Cursor, OpenClaw, Paperclip, Hermes, Instinct, and OpenWork ~ Codex and Claude Code session logs only (auto-discovery prefers Codex, --runtime claude selects Claude Code), plus support-billing CSV exports from six vendors ~ Claude Code and Codex, with the plugin install Claude Code-native 60+ formats auto-discovered (Claude Code, Codex, Gemini CLI, Copilot, Cursor, Zed, OpenCode, and more), plus opt-in web-chat imports since v0.46.0 (Claude.ai through a Chrome extension, ChatGPT, Codex web and cloud tasks) ~ about 13 named formats plus generic JSON/JSONL (Claude Code, Codex, Qwen Code, OpenCode, OpenClaw, Cursor exports, and more); Gemini CLI removed in v0.10.1 as discontinued upstream six named harnesses (Claude Code, Codex, Pi, oh-my-pi, OpenCode, Cursor) plus installed skills, local git history, and GitHub PRs with reviews and checks ingested into the same graph 18 named sources in the unified report (Claude Code, Codex, OpenCode, Amp, Droid, Codebuff, Hermes Agent, pi, Goose, OpenClaw, Kilo, Kimi, Qwen, Copilot CLI, Gemini CLI, Antigravity, Grok Build CLI, ZCode) ~ Claude Code and Codex through their native hooks plus transcript and rollout import, Cursor through compatible live events and ~/.cursor transcript discovery ~ Claude Code only (terminal, IDEs, anything writing ~/.claude) ~ 7 formats auto-detected (Claude Code, Cursor, Codex CLI, Gemini CLI, OpenCode via export, Kimi Code, Hermes Agent via SQLite or export) ~ OpenClaw only, through its eight-hook plugin ~ 42 integrations claimed (Claude Code, Codex, Cursor, Gemini, Amp, Command Code, and more); Cursor usage syncs from cursor.com with an app login since v0.9.26, replacing the local-estimate reads the launch thread criticized ~ Codex only, through one codex app-server per workspace; CLI sessions appear when their working directory matches the workspace ~ eleven named agents (Claude Code, Codex, Cursor, Gemini CLI, Antigravity, Copilot CLI, Muse Code, OpenCode, Amp, Hermes, Claude Desktop) through hooks, –agent flags, and plugins, plus any agent tagging a hook payload coucou_agent ~ about 40 agent harnesses documented (Claude Code, Codex, Cursor, Pi, OpenCode, and more) ~ 14 named agents through the Git AI standard (Claude Code, Codex, Cursor, Copilot, OpenCode, Pi, Windsurf, Droid, Amp, Gemini, Continue, Junie, Rovo Dev, Firebender) plus a documented add-your-agent path, hooks managed centrally by the CLI ~ 19 engines graded per capability (Claude Code, Codex, Cursor, OpenCode, Pi, GitHub Copilot CLI, Grok, Hermes, Kimi Code, KiloCode CLI, Goose, and more), support uneven ~ 46-plus tools graded per capability (Claude Code, Codex, OpenCode, Cursor, Antigravity, Copilot, and more); several limits sources are account-level cookie reads (Ollama Cloud, Alibaba Cloud), not session files ~ about 50 named local sources (OpenCode, Claude Code, OpenClaw, Codex, Copilot CLI, Gemini CLI, Cursor, Amp, Droid, Pi, Kimi, Qwen, Cline, Goose, Crush, Zed, Kiro, Devin, Junie, and more), several through cached account exports or estimation
Token cost reporting ✗ product events, not agent token cost ✗ computes no costs, it verifies whether records are countable (duplicates, re-emitted events, response conflicts); session and invoice totals are left null as UNPROVABLE ✓ per-session token usage and cost breakdowns (since v0.9.7) ✓ per-model pricing catalog, seconds over months of sessions ✓ per-session tokens and estimated USD cost across sources, with pricing overrides and confidence levels ✓ per-turn and per-session pricing from model pricing rows, queryable by session, text, commit, or branch, plus routing receipts (main-loop versus subagent split and routable-spend estimates) ✓ daily, weekly, monthly, and session reports plus Claude 5-hour blocks and a beta statusline, with per-model breakdowns; pricing-catalog estimates ✓ model-aware tokens with configurable per-model and per-session pricing, dated introductory rates, per-subagent cost, and compaction-safe totals ✓ per-turn token attribution across 7 context categories with compaction visualization, plus per-subagent tokens, duration, and cost ✗ ✓ per-step tokens and USD cost, 80+ models with cache-aware pricing ✓ by tool, model, project, git branch, and task, with pull-request cost views since v0.9.26 and plan-quota tracking; API-equivalent estimates marked with a tilde ~ sidebar account rate-limit and credits meter plus a home usage snapshot, no per-session or historical cost reports ~ plan-limit pills for Claude (5-hour and weekly) and Codex, no per-session token cost ✗ ~ tokens and cost aggregated per commit and PR across the sessions behind them (stats JSON), no per-session or per-day dashboards ~ per-engine token usage, off by default ✓ live tokens and cost in USD/TWD/HKD/CNY with daily exchange rates, cache hit versus miss breakdowns, CSV and JSON export, manual subscription cost records ✓ tokens and cost from LiteLLM’s pricing catalog with cache discounts, daily and per-model views
Search ~ flexible analytics queries (metrics, group_by, filters), funnels, paths, retention, not transcript search ✗ findings reports in JSON and HTML, no transcript search ✓ filtering and search across live and stored events ✓ FTS5 full text, semantic search opt-in ✓ sort and filter sessions by cost, duration, health, failures, anomalies, model, source, or text ✓ BM25 full-text recall across turns, commits, and skills, plus 31 read-only graph views ✗ report tables and JSON output, no transcript search ✓ searchable operational record across sessions, conversations, events, and workflows in the dashboard ✓ command-palette cross-session search, multi-pane side-by-side sessions ✗ chapter navigation and bookmarks inside a replay, no cross-session search ~ trace tree, call graph, timeline browse, and natural-language Ask Tracy queries over the trace graph, not transcript search ~ sessions drill down to turns and periods compare side by side, no full-text search ~ thread pin, rename, archive, and copy with per-thread drafts, no transcript search ~ Spotlight finds past turns on iPhone, no cross-session search ✓ cross-agent message and tool-call search, subagent and fork aware ✗ AI blame and stats lookups, no transcript search ✓ BM25 default, opt-in local or remote (OpenAI-compatible) embeddings for semantic and hybrid queries, saved memories searchable, SSH federation ~ per-session drilldowns, trend and heatmap dashboards, no full-text transcript search ✗ report tables and TUI views, no transcript search
Provenance ✗ ✗ ✗ ✗ ~ heuristic Git delivery correlation only, no line-level attribution ~ links sessions to commits and PRs and prices a branch’s sessions, no line-level attribution ✗ ✗ ✗ session-level inspection only, no code-line or commit attribution ✗ ✗ ~ spend matched to pull requests and git branches, not line-level attribution ✗ ✗ ✓ blame maps a line, file, commit, or PR to the session that produced it ✓ line-level attribution to agent, model, session, and prompt stored in refs/notes/ai, preserved through rebase, squash, stash, cherry-pick, amend, and reset; git mv, filter-branch, and git replace lose it ✗ ✗ ✗
Live observation ✓ real-time terminal dashboard across projects, plus opt-in web session replay ✗ retrospective reads of logs already written ✓ hook events stream to the dashboard over websockets as agents run ✗ retrospective only, parses files already written ✗ retrospective only, reads logs already written ~ live ingest triggers from the dashboard and stale reads quietly catch up in the background, but the surface is a published snapshot, not a hook stream ~ live Claude 5-hour block monitoring and a beta statusline, not a session dashboard ✓ hook and Cursor live-event streams render as turns happen, and the dashboard can launch, steer, resume, and reattach to agents ✗ retrospective, reads transcripts already on disk ~ --serve --watch follows a running session and renders the replay as it goes ✓ runs stream to the hosted dashboard through the eight-hook plugin ~ menu bar, tray, and Capacity Dock show current spend and what is running now, not agent actions ✓ the conversation view renders turns as the app-server produces them, with unread and running state, approval cards, and stop and interrupt controls ✓ hook and plugin events render as turns happen, approval cards wait in the notch, and Live Activities carry the session to the Lock Screen ✗ retrospective only ✗ checkpoints record while agents work, nothing renders live ✗ retrospective only ✓ widget updates within seconds of each turn, live tok/s rate, menu bar and tray popovers ✗ retrospective parsing
Team features ~ multi-agent access and cross-project portfolios, no team roles or SSO documented ✗ single-user local; the optional Actions workflow needs a separate private repository, and the README warns public forks are not private storage ✗ single-user local setup ✓ PostgreSQL push, machine-labeled sync, S3 roots, versioned exports ~ CI reports and shared baseline artifacts, no hosted team sync ~ a $20/seat founding-cohort team rollup over a private git repo the team owns (aggregates only) plus opt-in public profile gists; no hosted transcript storage ✗ ~ remote push ingestion for machines behind NAT, alerts with webhook fan-out, Docker/K8s/Terraform deployment docs, and optional auth tokens; no hosted team product ✗ single-user local, SSH inspection of remote machines ✗ the artifact itself is the sharing model, no team product ~ multi-tenant accounts (Tenant to Agent to Trace to Span) with referrals, no team roles documented ✗ ✗ single-user local, with an optional self-hosted remote daemon for running Codex on another machine ✗ single-user, sync is your own iCloud between your own devices ~ beta history sharing through a self-hosted ctx server with collection permissions and revocable access since v2.2.0, no hosted team product ~ open-source CI Actions preserve attribution through web-UI merges; the Git AI Platform adds MDM rollout, a hosted or self-hosted prompt store, and percent-AI by PR, repo, and team ✗ single-user, SSH federation to your own machines, object-storage sync announced ~ real-time multi-device sync through a self-hosted hub (in-widget, Node CLI, or Cloudflare Worker) or iCloud Drive, no team roles ✗ single-user; the opt-in leaderboard publishes public usage profiles
Privacy posture self-host keeps events in your D1 or SQLite, hosted cloud stores them in the vendor database, replay is opt-in and PII-masked local-first, nothing uploads by default; the free measurement runs receive only the sanitized sample you attach; reports hash response IDs with SHA256, which the guide warns can still link records across reports local-first, events stay in a local SQLite store behind a local Docker server local-first, one anonymous ping by default, disableable local-first, prompt and result bodies are not stored in tool steps, history of derived metrics is opt-in local-first, ingest makes no outbound calls; the only network surfaces are consent-gated opt-ins (public gists, profile widget, pattern PRs), and the team rollup ships redacted aggregates to your own repo local reads only, no accounts; fetches the model-pricing catalog unless run with –offline local-first, loopback binding by default, optional auth tokens when exposed beyond loopback, transcripts never sent to a hosted service local reads only, nothing leaves the machine; SSH reads are user-initiated local reads with automatic secret redaction before export; the web converter runs entirely client-side in the browser hosted, trace payloads including LLM inputs and outputs are uploaded to the vendor’s data lake local reads, no account; guard and optimize edit local config with backups, and the MCP server pseudonymizes project names local reads through your own Codex CLI login and config; remote mode rides your own Tailscale tailnet with a token you issue local-first, no telemetry, no account, keys in the system keychain; iPhone sync through your private CloudKit database with sensitive fields encrypted under your iCloud keys; the APNs relay sees only the agent name and state; the Android link rides your local network only with approvals bound to the exact pending request local-first, attribution refuses data not on the machine local-first, works offline with no login and no filewatchers, sessions are scanned, redacted, and stored outside git, attribution notes sync with the code through push and fetch local-first, no cloud service in the default path, embeddings local by default (the opt-in remote path sends transcript text to that API) local-first, per-session transcript reads happen on demand and are never synced, sync carries usage not transcripts local reads by default, tokscale submit uploads usage data to build a public leaderboard profile (opt-in)
Pricing free cloud tier (100k events/month, 2 projects), metered cloud at $1 per 10k events, self-host free free MIT CLI, public vendor rules, conformance pack, and claims index; the $10 and $20 paid diagnostics were retired 2026-10-21, replaced by free human-run measurement runs free, MIT, no paid tiers published free, MIT, no accounts free, MIT, no paid tiers published free AGPL-3.0 CLI; $20 per developer per month founding-cohort team plan (five-team cap); commercial licenses by contact free, MIT, sponsor-funded, no paid tier free, MIT, no paid tiers published free, MIT, the website lists a $0 offer free, MIT, no paid tier consumption credits; 100 free, packages $10 to $400, storage 1.35 credits/MB/day free, MIT, no paid tiers published free, MIT, no paid tier; the site still carries its beta disclaimer free, MIT, no paid tier; App Store builds coming free, Apache-2.0, the former $20/month pro add-on withdrawn from the site free, Apache-2.0 CLI; the teams platform is demo-gated with no published prices free, MIT, no paid tier published free, MIT, no paid tiers published, self-hosted sync free, MIT, sponsor-funded, no paid tier

Reading the matrix
#

The live-observation row is the row this matrix existed to name, and it now has six coding-agent answers: agents-observe streams hook events into its dashboard while the agents run, Claude Code Agent Monitor streams hooks plus Cursor live events and can also launch and resume agents from its dashboard, Coucou routes the same hook events into a notch island and onto the Lock Screen and is the only one that delivers the approval decision itself, ClawTrace fills the same row for OpenClaw by uploading the run, Token Monitor renders 46-plus tools’ usage and limits in one always-on widget as turns happen, and CodexMonitor renders Codex’s own app-server stream, approval cards, and thread state in a desktop command center, though it has been dormant since March 2026; Agent Analytics offers a real-time terminal dashboard for its own product events, though the hook-and-stream coding-agent cells still cover only Claude Code, Codex, Cursor, OpenClaw, and Coucou’s eleven, so the remaining empty cells belong to the harnesses none of them hook. The second row worth reading is provenance, which gained a second answer this pass: ctx blames from the session side, mapping a line, file, commit, or PR back to the transcript that produced it, while git-ai records authorship inside git as the agents work, so the answer sits in refs/notes/ai, syncs with the code, and outlives the transcript retention window that bounds every other member; agentsview still deliberately leaves both to cost and history questions.

Token cost reporting is the row that pays for the tool: harness-native cost views reset and see only their own sessions, while a pre-indexed store answers multi-tool, multi-month questions in seconds. An earlier version of the agentsview docs benchmarked its reports at 84 to 223 times faster than ad-hoc parsing (calling that an upper bound); the current docs have dropped that benchmark entirely, so the row should be read as “fast because pre-indexed”, with no vendor number left to lean on. The docs’ token-usage page documents schema version 5 again as of 2026-10-21 (it documented version 6 from 2026-10-10 through 2026-10-19), the sixth flip this section has recorded, so scripts consuming those reports should expect churn either way. AgentTrace adds a cost cell the archive tools do not, with explicit estimate labeling and pricing overrides, while ClawTrace adds per-step cost inside a hosted trace, which is more granular than anything local but only for OpenClaw. ccusage and CodeBurn make cost the whole product: ccusage is the zero-install incumbent the other tools benchmark against, CodeBurn adds task and branch cuts plus an optimize loop that proposes config fixes, and both print estimates computed from pricing catalogs rather than bills. Tokscale and Token Monitor extend the same estimate approach to about 50 and 46-plus sources respectively, Tokscale on LiteLLM’s catalog and Token Monitor adding currency choice and manual subscription cost records. AgentMeasure sits apart from the row on purpose: its September 2026 audit of about 110 usage tools ended in merged double-count fixes in the CodeBurn, Tokscale, and Token Monitor columns, so every estimate cell above now carries an externally probed error history, not just each tool’s own caveat. AgentMeasure itself went entirely free on 2026-10-21, retiring its $10 and $20 paid diagnostics in favor of free human-run measurement runs.

Memex adds the closing move the other archive tools lack: resume-in-place, where finding a session and re-entering it are one action in the TUI, plus the category’s only semantic search (BM25 by default, opt-in embeddings that run locally or, since 2026-10-03, against an OpenAI-compatible API that receives your transcript text, so the headline requires setup either way). Its 19-engine support table is graded per capability and unevenly at that (some engines have no resume at all, token counting is missing for at least one), and its launch footprint is as thin as ctx’s was, so read the column as promising and unproven.

The 2026-09-27 columns split the category’s remaining questions: AgentTrace is the local cost, latency, and failure auditor that adds slow-run diagnosis the archive tools skip, ClawTrace is the hosted answer that trades privacy for full LLM payloads and an AI analyst, and Agent Analytics tests where the category’s boundary sits by serving product analytics to agents instead of session data.

The 2026-10-06 columns push on three remaining fronts: Tokscale widens the ledger to about 50 declared sources and adds the category’s first opt-in public leaderboard, Token Monitor makes the widget live and multi-device and keeps history after harnesses prune their transcripts, and claude-devtools shows how deep one harness’s logs go when the terminal stops telling you, per-turn token attribution across seven context categories included.

claude-replay adds the output format the rest of the category never produces: a single dependency-free HTML file you can email, host, or embed, compiled from seven harnesses’ transcripts with secret redaction applied, which makes it the only member whose reader is someone who never installed the tool, and its ledger cells (cost, search, provenance) are empty by design rather than by gap.

Breadth of coverage is agentsview’s moat: roughly 60 supported sources against Tokscale’s about 50, Token Monitor’s 46-plus, CodeBurn’s claimed 42, ctx’s 40, ccusage’s 18, Memex’s 19, AgentTrace’s 13, AgentMeasure’s two harnesses, git-ai’s 14 named agents, Coucou’s eleven named agents, ax’s six, claude-devtools’ single harness, the two harnesses agents-observe covers, the three Claude Code Agent Monitor watches, or the one each that ClawTrace hooks and CodexMonitor drives, which matters because most practitioners now run two or three harnesses at once. Agent Analytics is agent-agnostic by design, since any agent that can call HTTP is a client. ctx’s moat is different: agent-facing retrieval, where the consumer of the search is your next session rather than you.

Choosing from the matrix
#

  • Running three or more different coding agents and wanting one private history and cost view: agentsview.
  • Wanting your agents to recall why code exists, from the session that wrote it: ctx, with its self-reported numbers accepted.
  • Wanting to know which agent, model, and prompt wrote a line of code, with the answer stored in the repo itself: git-ai, accepting that only agents following its checkpoint standard are attributed.
  • Living in the terminal and wanting semantic recall plus one action back into the session: Memex, accepting a few hundred stars and thin verification.
  • Wanting your history to fix your next run, with recall, branch costs, skill scores, and model routing in one local graph: ax, accepting a 121-star, AGPL-licensed, solo-maintained 0.x.
  • Auditing what a run cost and why it was slow, locally and with CI gates: AgentTrace.
  • Needing a usage or billing number to survive a dispute before leaning on it: AgentMeasure, which audits the records and marks what cannot be proven rather than printing an estimate.
  • Wanting the quickest terminal answer to what did my agents cost this week, without installing anything: ccusage.
  • Wanting spend cut by task and branch, plan-quota tracking, and a guard that stops a runaway session: CodeBurn.
  • Wanting the widest declared token ledger, and maybe a public profile of your usage: Tokscale, after reading its submit flow first.
  • Wanting one live desktop view of tokens and limits across every tool, synced across your machines: Token Monitor, accepting a fast churn rate.
  • Wanting to know why a Claude Code context window filled and exactly what a tool call did: claude-devtools.
  • Wanting one live control room over Claude Code, Cursor, and Codex that also imports their history and can launch or resume agents: Claude Code Agent Monitor, accepting a single-maintainer, many-featured surface.
  • Running Codex and wanting one desktop command center over its threads, approvals, and account usage: CodexMonitor, accepting seven months without a commit while the vendor’s own app absorbed the job.
  • Wanting approvals to follow you away from the desk, to the notch, a locked iPhone, or (with the Windows and Linux apps) a linked Android phone: Coucou, accepting a weeks-old 0.x app with no independent review yet.
  • Running OpenClaw and wanting full-payload traces plus an AI analyst, hosted: ClawTrace, accepting that the traces leave your machine.
  • Instrumenting a product your agents are building and reading its traffic through an agent: Agent Analytics.
  • Needing to show a session to someone who has no tooling installed: claude-replay, compiling the transcript into one HTML file they can open anywhere.
  • Needing to see what an agent is doing right now: agents-observe for Claude Code or Codex, Claude Code Agent Monitor to add Cursor and to drive agents from the same view, or ClawTrace if you run OpenClaw and can upload the run; otherwise a harness-native view is still the fallback.
  • Single-agent users: your harness’s built-in usage views are probably enough.

Changes
#

  • 2026-08-30 - Created as the Session analytics category’s companion matrix, a single-column scaffold with the live-observation gap named in prose.
  • 2026-09-05 - Extended from one to two columns with ctx.
  • 2026-09-06 - Updated the ctx cell for its published pro pricing.
  • 2026-09-08 - Dropped the agentsview benchmark claim from the prose after the tool’s docs removed it entirely.
  • 2026-09-13 - Added the agentsview usage-output schema version 6 churn note to the reading prose.
  • 2026-09-16 - Extended to three columns with agents-observe, filling the live-observation gap the prose used to name.
  • 2026-09-18 - ctx agents-covered cell updated after the docs began listing about 40 supported agent harnesses.
  • 2026-09-20 - Extended from three to four columns with Memex, inserted last alphabetically, every row gaining a cell traced to the new note, with the reading and choosing prose extended.
  • 2026-09-20 - Corrected the schema caution in the reading prose: the docs’ usage JSON contract is at schema version 5, version 6 is the session-export schema.
  • 2026-09-24 - Removed the verification preamble line on owner request.
  • 2026-09-25 - Retired the ctx pro cells: the Kind, Open source, Team features, and Pricing cells now describe blame as built in, after v2.0.0 shipped it inside the single open-source executable and the site dropped every paid-tier listing.
  • 2026-09-27 - Re-verified all four columns: repository counts refreshed (agents-observe 684, agentsview 6,001, ctx 1,135, Memex 231 stars), the reading prose’s schema line revised after the agentsview docs returned to version 5, and the Memex star count in the choosing prose generalized; no table cells moved.
  • 2026-09-27 - Extended from four to seven columns with Agent Analytics, AgentTrace, and ClawTrace, re-sorted all columns alphabetically, updated the intro member count and framing, and extended the reading and choosing prose.
  • 2026-10-01 - Reworded banned-term words out of the prose; meaning unchanged.
  • 2026-10-02 - Moved the ctx Team features cell after v2.2.0 shipped opt-in history backup plus beta sharing through a self-hosted ctx server; the other six columns re-verified unchanged.
  • 2026-10-03 - Updated the Memex agents-covered cell and the prose engine counts to 17 engines after v0.25.0 (2026-10-02) added Hermes and KiloCode CLI; the other six columns re-verified unchanged.
  • 2026-10-04 - Updated the Memex Search and Privacy posture cells and the reading prose after v0.26.0 (2026-10-03) added opt-in OpenAI-compatible remote embeddings alongside the local path; the other six columns re-verified unchanged.
  • 2026-10-05 - Extended from seven to nine columns with ccusage (the zero-install cost-report incumbent, inserted in sorted position between AgentTrace and ClawTrace) and CodeBurn (the desktop spend-analysis app with task attribution and spend guards, between ClawTrace and ctx), every row gaining two cells traced to the new notes; the AgentTrace Deployment cell dropped winget after v0.10.0 removed it, the intro count and reading and choosing prose updated.
  • 2026-10-06 - Extended from nine to twelve columns with claude-devtools (the single-harness Claude Code session DevTools), Token Monitor (the live desktop usage widget with provider limits and self-hosted multi-device sync), and Tokscale (the about-50-source token ledger with an opt-in public leaderboard), inserted in sorted position, every row gaining three cells traced to the new notes; the AgentTrace agents-covered cell dropped Gemini CLI after v0.10.1 removed it, the intro count and the reading and choosing prose updated, and the agentsview schema note moved back to version 6 per the current docs.
  • 2026-10-07 - The reading prose’s schema line moved back to version 5 after the agentsview docs flipped again (version 6 documented on 2026-10-06, version 5 on 2026-10-07), the fourth recorded flip; no table cells changed.
  • 2026-10-07 - Extended from twelve to thirteen columns with claude-replay (the transcript-to-replay compiler producing self-contained HTML artifacts across seven harnesses), inserted in sorted position between claude-devtools and ClawTrace and traced to the new note; the intro reframed the category as four arrangements with the shareable artifact as the new one, and the reading and choosing prose gained the claude-replay entries.
  • 2026-10-10 - Extended from thirteen to fourteen columns with Claude Code Agent Monitor (the local live control room over Claude Code, Cursor, and Codex, native hooks plus transcript discovery, with launch-and-resume control), inserted in sorted position between ccusage and claude-devtools and traced to the new note; the intro, reading, and choosing prose gained its entries, and the agentsview schema line moved back to version 6 per the current docs (the fifth recorded flip).
  • 2026-10-10 - Extended from fourteen to fifteen columns with Coucou (the notch-and-iPhone companion watching eleven named agents live, with approvals from the notch or Lock Screen through your own iCloud), inserted in sorted position between CodeBurn and ctx and traced to the new note; the intro’s live-view list, the live-observation reading (four answers became five), the coverage moat line, and the choosing prose gained its entries.
  • 2026-10-11 - Updated the CodeBurn column after v0.9.26 and v0.9.27: the Deployment cell gained the VS Code extension and the one-click updater, the agents-covered cell moved to 42 claimed integrations with Cursor usage now syncing from cursor.com since v0.9.26, and the token-cost cell gained pull-request cost views and the tilde labeling on estimated dollars; the breadth line moved from 41 to 42, and the other fourteen columns re-verified unchanged.
  • 2026-10-11 - Extended from fifteen to sixteen columns with git-ai (the Apache-2.0 git-layer provenance recorder, joined into OpenAI’s Codex team on 2026-09-11, whose line-level attribution lives in Git Notes and survives every rebase), inserted in sorted position between ctx and Memex and traced to the new note; the intro reframed the category as five arrangements with git-layer provenance as the new one, the founding-question and provenance readings gained its entries, the coverage moat line moved to fourteen named agents, and the choosing prose gained a git-ai bullet.
  • 2026-10-13 - Updated the Coucou column after Coucou for Android 0.1.0 and the v0.4.0 Windows and Linux builds: the Kind cell gained the Android app beside the iPhone, the Deployment cell gained the Android app and its local-network pairing with the Windows and Linux apps, and the Privacy posture cell notes the local-network-only link, while the intro and choosing prose name the Android phone; the other fifteen columns re-verified unchanged against their notes.
  • 2026-10-16 - Extended from sixteen to seventeen columns with AgentMeasure (the local evidence-discipline auditor whose September 110-tool audit produced merged double-count fixes in the CodeBurn, Tokscale, and Token Monitor columns), inserted in sorted position after Agent Analytics and traced to the new note; the Memex agents-covered cell moved to eighteen engines after v0.29.0, the intro, reading, and choosing prose gained its entries, and the breadth line moved Memex to eighteen and named AgentMeasure’s two.
  • 2026-10-21 - The reading prose’s schema line moved back to version 5 after the agentsview docs flipped again (version 6 documented from 2026-10-10 through 2026-10-19, version 5 re-verified 2026-10-21), the sixth recorded flip; no table cells changed.
  • 2026-10-23 - Extended from seventeen to nineteen columns with ax (the local agent-experience graph ingesting six harnesses plus skills, git history, and GitHub PRs into an embedded DuckDB graph, with recall, cost routing, and the retro improve loop, inserted after AgentTrace) and CodexMonitor (the dormant MIT Tauri command center for Codex, last commit 2026-03-26, inserted after CodeBurn), every row gaining two cells traced to the new notes; the intro, reading, and choosing prose gained their entries.
  • 2026-10-23 - Volatile cell and prose updates: the agentsview agents-covered cell gained the v0.46.0 opt-in web-chat imports, the Memex agents-covered cell moved to nineteen engines after v0.31.0 added Goose, the Token Monitor Kind and agents-covered cells moved to 46-plus tools, and the AgentMeasure Pricing and Privacy posture cells recorded the 2026-10-21 retirement of its paid diagnostics; the reading prose counts moved with them.

See also
#

References
#