What did your coding agent actually do today?
agentlog reads your local Claude Code and Codex session logs and answers that in a screen. It has two faces:
- At a terminal it writes a standup-style brief: what each project was about, whether it got done, and what is waiting on you. This asks a model CLI already on your machine —
claudeorcodex, whichever is hosting the shell you run it in, or whichever is installed — and it is the one thing here that sends your day anywhere — see Privacy. - Everywhere else — pipes, hooks, cron, or
--log— it prints the plain log: which projects, which files, what failed. Offline, no network, no API key.
Everything is organised by project, because that is the first thing you want to know. Sessions, IDs and token counts are still there — one flag away — but they are not the headline.
--html writes the same digest as a self-contained file you can drop in a chat or share with a teammate. It works offline, requires no server, and contains no external assets.
pip install stillworks # one install, all five agent tools, including this one
agentlog today
# Or run from a checkout, no install needed — it is stdlib only:
cd /path/to/agentlog && python3 -m agentlog today(Since stillworks 0.2.0 the whole family ships in one wheel. The command, the
module, and the repo are all just agentlog; the old standalone PyPI name
agentlog-tool is the 0.1.x era and gets no further releases.)
What the log looks like (agentlog --log on 2026-08-03 — an invented day;
yours shows your own prompts and files. The brief, the terminal's default
face, is shown under "View flags" below):
2h 58m active across 3 projects · on 2026-08-03
shop-api 1h 41m 21 files edited · 214 commands · 9 errors
asked "add a retry with backoff to the payment webhook, and make the
failure path land in the dead-letter table so nothing is…"
edited …/webhooks/payments.py, …/dead_letter.py, …/test_webhooks.py
failed pytest tests/test_webhooks.py -k retry …
alembic upgrade head …
docs-site 49m 6 files edited · 78 commands
asked "rewrite the quickstart so someone who has never seen the CLI
can follow it — every command copy-pasteable…"
edited docs/quickstart.md, docs/install.md, mkdocs.yml
infra 28m 3 files edited · 51 commands · 2 errors
asked "why did the staging deploy roll back last night? read the
pipeline logs and tell me before touching anything…"
edited …/deploy/staging.yml, …/deploy/rollback.sh
failed kubectl logs deploy/api -n staging --since=12h …
5 sessions · 3 claude, 2 codex · busiest 15:00–16:00
compacted 17x in 2 sessions · 1h 3m spent on it, 2,481,506 tokens dropped
more: agentlog list · agentlog show ID · agentlog --sessions
Busiest project first. asked is the prompt you typed that names the work — the
one row that says what the session was for; the files are the ones written
most often, and failed names the command behind each error rather than just
counting them.
Generate an HTML digest you can share:
agentlog today --html today.html
# then open today.html in a browser — no server requiredagentlog same as: agentlog today
agentlog today | yesterday | week
agentlog since WHEN 10m, 2h, 3d, 1w, or a date like 2026-08-03
agentlog on DAY one whole day: 2026-07-31, or 3d for three
days ago
agentlog show SESSION_ID one session in full detail
agentlog list 50 most-recent sessions as a compact table
agentlog list --all all sessions (no row limit)
agentlog list --limit N show at most N sessions
agentlog goal "TEXT" declare what this project's work is for;
replayed at every resume from compaction
agentlog goal show the declared goal; --clear forgets it
agentlog handover run by an agent hook, not by you — see
"The note a session leaves itself"
agentlog list (first 3 rows):
ID PROJECT WHEN DUR SRC
-------- ------------------------ ---------------- -------- ------
019fc4b9 shop-api 2026-08-02 23:05 6m 18s codex
019fc4a7 shop-api 2026-08-02 22:45 6m 41s codex
019fc4a1 docs-site 2026-08-02 22:39 4m 29s codex
View flags:
--sessions the old per-session view: one block per session, with IDs,
models, turn counts and token totals
--brief a standup report instead of a log: what each project was
about, whether it got done, and what is waiting on you.
This one asks a model, so it sends the day off this machine
— see Privacy. At a terminal with a model CLI installed
(claude or codex) it is already the default; the flag
forces it anywhere else
--log the plain per-project digest, no model asked, nothing
leaves this machine — the way to say "not the brief" at a
terminal
--project NAME only projects whose name or path contains NAME
--file PATH read only this transcript, all of it (the time command does
not apply)
--file is for the caller that already knows which file it wants. Reading one
transcript takes about two seconds where reading a whole home takes minutes,
which is the difference between fitting inside an agent hook's timeout and not.
The time command is ignored on purpose: a session you name by path is reported
whole, including the part of it that happened yesterday. Which agent wrote the
file is worked out by reading it, not from where it sits or what it is called,
so a transcript copied somewhere else — or truncated so its opening records are
gone — still reads.
The brief is the answer to "what did you get done today?", which is a
different question from "what happened today?". It reads the day's
transcripts, works out what each project was about, and judges whether it got
done. Declaring a goal with agentlog goal is optional; where one exists the
project is judged against it and nothing else (a synthetic day):
today, Sat 8 Aug · 32m 00s of agent time · 2 goals in play
◐ billing — moving invoices to the new PDF library
goal: "Move invoices to the new PDF library; done when goldens match." —
the goldens still diverge on discounted invoices and the run ended red.
1 session · 32m 00s · 13k tokens · 1 file edited · 1 error
✔ shop-api — payment webhook retries
inferred goal: stop checkout silently dropping failed payments — retries
with backoff went in, committed and pushed with the suite green.
1 session · 30m 00s · 9k tokens · 1 file edited
✔ docs-site — quickstart rewrite
inferred goal: make the first page somebody lands on current — the
quickstart was rewritten and committed.
1 session · 15m 00s · 3k tokens · 1 file edited
Plus 2 sessions across 2 projects, 5m 00s total — nothing finished there.
○ not started — importer — "Rewrite the importer. Why: malformed rows are dr…"
Waiting on you
billing — decide whether discounts round half-up or half-even.
──────────────────────────────────────────────────────────────────────────────
5 sessions · 3 claude, 2 codex
3 files edited · 7 commands · 1 error · 26k tokens
Each project opens with a heading line — its name and, in a few words, what
the work was about — so the page reads in a couple of seconds. A goal you
declared is quoted on the next line, word for word; where none was declared
the model says what it understood the goal to be, labelled inferred goal: so
the two are never confused.
The prose is written by a model; nothing else is. Every figure on the page is computed from the transcripts, a sentence in which the model states a tally is dropped before it can contradict them, and a project it invents is dropped whole. The marks are earned, not claimed: ✔ done needs a receipt in the transcripts — a commit, a push, a release — or the claim is printed as ◐ in progress, which is the most it can prove; ✘ failed is the model's verdict; ○ not started is read off the goal store with no model in the room — a declared goal whose directory shows up in none of the day's transcripts. A project the model could not call one way or the other gets the undecided mark ·. Waiting on you collects what is blocked on a decision of yours; with no model installed the page degrades to the facts — goals, receipts, what was still failing — and says which part is missing.
The same question asked twice is answered from a small cache
(~/.agentlog/answers.json), so running agentlog again over an unchanged
day is instant and free; new work on disk changes the question and the cache
steps aside. The model asked is the one hosting the shell: pressed from
inside a Codex session the brief is written by codex, from inside Claude
Code by claude, and at a bare terminal by whichever of the two is
installed. Each is spoken to in its own dialect (claude -p on stdin;
codex exec --ephemeral, sandboxed read-only so the ask itself never lands
in the session logs agentlog reads), and each model's answers are cached
separately. AGENTLOG_MODEL_CMD names a different command to ask instead;
it is given the prompt on standard input.
Output flags:
--html FILE write a self-contained HTML digest to FILE
(time commands only: today, yesterday, week, since, on)
--md [FILE] Markdown to FILE, or stdout if FILE is omitted
(time commands only)
--json JSON to stdout; works with all commands including list and show
--verbose show parsing diagnostics (skipped-line counts)
--home DIR override home directory; used by tests and CI
The AGENTLOG_HOME environment variable is equivalent to --home.
Exit codes: 0 normal, 2 usage or argument error, 130 stopped by ctrl-c,
141 the reader hung up (agentlog today | head, or | less quit with
q). The last two are deliberately not 0: a digest that was cut off
short reported nothing about your day, and agentlog today > digest.md && mail-it should not mail half of one.
A long conversation gets compacted: the transcript is replaced by a summary so the work can carry on, and what the agent knew about the last four hours becomes whatever the summary kept. The transcript is still on disk and still complete, so the facts are not gone — only the agent's hold on them is.
agentlog handover reads the transcript in the moment before compaction, writes
down what the session had actually done, and hands it back when the session
resumes. Two hooks in ~/.claude/settings.json, the same line twice:
{
"hooks": {
"PreCompact": [
{"matcher": "auto", "hooks": [{"type": "command", "command": "agentlog handover || true"}]},
{"matcher": "manual", "hooks": [{"type": "command", "command": "agentlog handover || true"}]}
],
"SessionStart": [
{"matcher": "compact", "hooks": [{"type": "command", "command": "agentlog handover || true"}]}
]
}
}Which of the two jobs to do is read off the event name in the payload, so there
is no flag to get the wrong way round. The || true is not superstition: a
PreCompact hook that exits non-zero blocks the compaction it was watching, and
the one way this command can exit non-zero is by being an older agentlog that
has never heard of handover. One idiom and an upgrade half-done can no longer
stop your session. What comes back is the same digest the
plain command prints — where the work was, how long, which files were edited,
which commands kept failing — under a line saying where it came from.
Three deliberate choices:
- No model. Compaction already writes a summary; a second account of the same conversation would drift the same way. What a compacted session cannot reconstruct is the plain record, and that is all this hands over.
- Once. The note is deleted as it is handed over. Left behind it would be injected again at the next compaction, stating two-hour-old facts in the present tense.
- Never in the way. Every failure exits 0 and explains itself on stderr. A
PreCompacthook that exits non-zero blocks the compaction it was watching, so the agent stops and the reason is the note-taker.
The note carries receipts — directories, files, failing commands — and receipts cannot say what the work was for. That is what drifts: each summary of a summary reframes the job a little, until the session is confidently doing something adjacent to what was asked. Nothing in a transcript is reliably "the goal" — the first prompt is history, the latest is a correction — so the goal is declared, in so many words:
agentlog goal "Ship the importer. Done when a malformed row is reported and the clean rows still land. Constraint: no new dependencies."
The convention: declare it when a brief is accepted — by you, or by the agent
the moment it takes the job — redeclare it when the brief changes, and
agentlog goal --clear when it is done. From then on every resume from
compaction opens with the declaration, quoted verbatim under a label saying
when it was set and that newer instructions from the user override it. A
quotation cannot drift; the label keeps it from overruling a legitimate pivot.
Say the objective, what done looks like, and the constraints that bind it —
not the plan. A plan changes weekly, and a stale plan replayed as current is
the drift this exists to prevent; the plan's residue is already in the note.
If the goal has a reason worth keeping in front of the work, give it after
Why: — agentlog goal "Rewrite the importer. Why: malformed rows crash it." — and --brief quotes that reason instead of inventing one.
There is a hard cap (2,000 characters) and a declaration over it is refused
with the count, so the anchor can never become the bloat.
Goals are keyed by the directory they are declared in — and by the session,
when the declaring shell belongs to one. Declared inside a Claude Code
session (the agent accepting a brief), the goal binds to that session and is
never replayed into any other: two sessions working the same directory are
two briefs, and crossing them would be the drift this exists to prevent.
Declared at a bare terminal, it is the directory's shared north star — the
fallback any session without a goal of its own resumes with. Either way it
lives in the same private ~/.agentlog store as the notes — never in your
repository.
It is quick enough to sit in a hook's timeout: about two seconds on a thirty-hour session, against minutes for a whole-home scan.
agentlog reads JSONL files in ~/.claude/projects/**/*.jsonl (Claude Code)
and ~/.codex/sessions/**/*.jsonl (Codex). It never writes to those files
or uploads anything.
For each session it derives:
| Shown | Derived from |
|---|---|
| project | cwd field in first user record |
| start / end / duration | first and last timestamp fields seen |
| models | message.model in assistant records |
| user turns | type == "user" records that are a person typing — not tool results, not subagent prompts (see below) |
| files read | Read tool-use calls (input.file_path); for Codex, the paths handed to a reading command in a custom_tool_call snippet or an older exec_command — Codex has no read tool, so a read is a cat or a sed -n and the command text is the only record of it |
| files written | Write, Edit, MultiEdit tool-use calls (input.file_path) and NotebookEdit (input.notebook_path); Codex patch_apply_end records, plus *** Update File: lines inside older apply_patch envelopes |
| commands | Bash tool-use calls (input.command); Codex custom_tool_call script snippets, plus older exec_command and apply_patch calls |
| errors | tool_result records with is_error: true; Codex command output with a non-zero exit code, a patch that would not apply, and an mcp_tool_call_end whose result is an Err |
| the failing command | the tool-use call the failed result points back at (tool_use_id / call_id) |
| recap | system records with subtype away_summary — the short plain-English note a background session writes at the end of a turn, minus its trailing pointer at the settings screen |
| context | system records with subtype compact_boundary — how many times the session ran out of room and summarised itself, how long that took (compactMetadata.durationMs), and how much was thrown away (preTokens - postTokens, see below) |
| tokens | message.usage.input_tokens, cache_creation_input_tokens, and cache_read_input_tokens in assistant records; Codex uses the final total_token_usage snapshot, falling back to summing the per-turn last_token_usage blocks in a log too old to carry it |
A turn is a time you said something. Claude Code writes a type: "user"
record for four different things, and only one of them is a person: a tool
result fed back into the loop is the agent's own machinery, a record marked
isSidechain: true is a prompt the agent wrote for a subagent, and a record
marked isMeta: true is Claude Code putting text into the conversation on its
own account — the caveat that precedes a slash command's output, the body of a
skill being loaded, a message relayed from another session, a nudge to
continue, the placeholder standing in for a pasted image. Counting all four
reported 38318 turns across 896 real session logs where 2314 were typed, and
one session read 3637 for a day with 211. That kind of error is worth more
care than most, because a reader cannot catch it: an under-count can be checked
by adding up the sources, but there is nothing to check an over-count against —
it just reads as a long day. The subagent's work still counts; a pytest -x
it ran is a command that ran, and so does everything that happened around an
injected record. Only the claim that you spoke is dropped. A record with
neither field — older logs — is counted, since dropping real turns would be the
opposite mistake, and only an explicit true counts as either.
Compaction is where long sessions go. When a session runs out of room,
Claude Code summarises the conversation so far and throws the rest away. It
costs real wall-clock and it is not free of consequence, but nothing in the log
a person reads says it happened, so a session that spent hours re-reading its
own summary looks exactly like one that was merely slow. agentlog show now
prints a context line when there was one:
context compacted 98x — 3h 59m spent on it, 9,944,222 tokens dropped
The digest says it too, because show needs a session id and a session id
needs you to already suspect which session to look at:
compacted 127x in 6 sessions · 4h 54m spent on it, 13,239,126 tokens dropped
The number of sessions is in that line on purpose. Twelve compactions in one session is a session that should have been split; twelve across twelve sessions is an ordinary week, and the count alone cannot tell them apart.
Across the 896 logs this was developed against: 313 compactions in 49 sessions,
twelve hours of wall-clock, a median of 2m17s each, and a median of 13% of the
context surviving. Manual /compact runs are counted separately from
automatic ones — one is a person deciding, the other is the session hitting a
wall, and a number that mixed them would overstate how often it hit the wall.
The subtraction matters. The record also carries cumulativeDroppedTokens,
which is a running total — it equals the running sum of preTokens - postTokens on every real record — so adding that field up across a session
counts the first compaction once for every compaction after it. A session with
three of them would report roughly three times what was lost, and the only
thing wrong with the number is that it is too big, which is not something a
reader can check. agentlog uses pre - post per compaction and lets the total
fall out of the sum.
Tool-use IDs are deduplicated so streaming-split records are not double-counted.
Malformed lines are skipped silently; their count appears under --verbose.
Codex has sent its work in two different shapes. Older sessions announce each
command as a function_call named exec_command, with the command in a
structured field. Current builds send a custom_tool_call instead: the record
carries a snippet of JavaScript — tools.exec_command({cmd:"pytest -x", ...}),
often several inside one Promise.all — and the command has to be read back out
of the source. Both are parsed. Until v0.2.3 only the first was, which on the
machine this was found on made 65.6% of the recorded Codex work invisible and
left 74% of Codex sessions looking like the agent had done nothing; for the
current month it was 98%. Nothing failed — the sessions were listed, with no
commands and no files under them, which is exactly what a session of pure
conversation looks like.
MCP tool calls are reported in their own record, and only a failed one is shown.
A successful MCP call is not turned into a command — it is not a shell command,
and the command list means one thing — so a session that spent its time in MCP
tools will look quiet. Its failures will not: before v0.2.3 they were in a
record nothing read, and four sessions on the logs this was found with reported
0 errors while an MCP server they had asked for was not running.
The read/written split is specific to Claude Code. Codex has no structured
file-write field — it edits by handing a patch envelope to a patch tool — so
its written files come from the patch_apply_end record, which names them
absolutely and is also the only place a patch that failed is admitted. A
patch that did not apply is counted as an error, never as a write.
Some builds never send that record: they hand the envelope to a
custom_tool_call — sometimes bare, sometimes wrapped in a JavaScript string
literal with the newlines escaped — and report nothing afterwards. A session
that emitted no patch_apply_end at all therefore falls back to the
*** Update File: / *** Add File: / *** Delete File: lines in the envelope
itself. The fallback is decided per session, not per call: if the session sent
even one end record, that build does report its own patches and the envelopes
are ignored, because the end record knows which patches failed and the envelope
does not. Listing an edit that never reached the disk is the worse error of the
two — nothing else in the report contradicts it. Relative paths are resolved
against the call's working directory, so one file is not counted twice under two
spellings. Codex files it only reads are not distinguishable from any other
shell command, and are not reported.
You spent the day with Claude Code or Codex. What actually happened?
- Which files did it touch?
- How many commands did it run?
- Which sessions ran long?
- Where did things break?
That information is in your JSONL session logs, but reading raw JSONL is not practical. agentlog parses it and gives you a summary you can read in seconds.
The HTML output (--html) is meant to be shared: a single offline file
containing the activity metadata of one or more sessions, useful for stand-ups,
reviews, or just keeping your own notes.
Why not just ask my AI to summarize the logs? That works, but it involves sending your session logs to a model API. agentlog stays entirely local: it reads log structure, not content, and nothing leaves your machine.
Many tools read Claude Code JSONL logs. agentlog occupies a specific gap.
ccusage (ccusage.com) Token and cost tracking for Claude Code, Codex, and 15+ other agent CLIs. Daily/weekly/monthly summaries, per-project filtering, JSON output, MCP server. It reads the same files agentlog reads. Session data includes aggregate tool-call counts but surfaces them as numbers, not as enumerated file and command lists. agentlog is complementary: ccusage answers "what did it cost?" agentlog answers "what did it do?"
claude-code-log Python CLI that converts Claude Code JSONL transcripts to self-contained HTML or Markdown. Excellent for reading full conversation content. Works per-file or per-project. agentlog differs in focus: it shows metadata (files, commands, durations) rather than conversation content, and aggregates across multiple sessions in a time range.
claude-code-trace JSONL session viewer shipped as a native desktop GUI, web app, and TUI. Shows conversations, tool calls, and token counts per session with live tailing. No standalone HTML export and no cross-session time-range digest.
Claudoscope Native macOS menu bar app. Session browser, cost/token stats, generated-files view, full transcript, real-time secret scanning. macOS only, no CLI mode, no HTML export, not scriptable.
CASS — coding_agent_session_search Unified TUI and CLI for indexing and searching session history across 11+ agent providers using BM25 and semantic embeddings. SQLite-backed. A search tool, not a digest generator.
vibe-log-cli Daily activity summary for Claude Code and Codex sessions. Analysis is routed through ACP — it uses your local Claude Code or Codex instance to produce the summary rather than parsing JSONL structure directly. Requires a running agent.
agentlogs (agentlogs.ai) Cloud-based team observability: captures and uploads session transcripts. Requires login and network upload.
agent-sessions Local-first macOS app for browsing, searching, and resuming sessions across many providers. macOS only, no CLI, no HTML export.
claude-session-analyzer Quantitative behavioral analysis: thinking depth, Read/Edit ratio, rework indicators. Aimed at workflow optimization, not readable activity digests.
Where agentlog fits:
agentlog list and agentlog show are the scriptable, terminal-native way to
browse sessions without launching a GUI or browser. agentlog --html produces
a self-contained offline file for sharing — the existing GUI tools are
session-by-session viewers, not cross-session digest generators.
Schema drift. Claude Code and Codex change their log formats without
notice. Fields that agentlog reads today may move or disappear. Parsing is
defensive and will not crash, but some sessions may show empty file or command
lists if the format changes. That has already happened once, and is the worst
failure this tool has: an empty list is indistinguishable from a quiet session,
so a whole format going unread looks like the agent was idle rather than like a
bug. Both known Codex shapes are parsed side by side rather than one replacing
the other, and --verbose names files that could not be read at all.
A Codex read is a guess made from a command, and it under-reports on
purpose. Claude Code reads a file by calling a tool named Read, so the path
is a field. Codex has no read tool: it reads by running sed -n '1,200p' notes.md, and the command text is the only record. Every Codex session
therefore reported no files read until v0.2.5 — not an error, just a zero that
looked like a fact.
They are read out of the command now, by a rule built to miss rather than to
invent: a path is counted only when the verb opens everything it is handed
(cat, head, tail, sed, wc, md5sum and a dozen more). Nothing that
searches is counted at all — rg pattern src/ puts a pattern, a glob and a
directory in the position a path goes, and the text cannot say which is which,
so grep and rg reads are missed on purpose. A file named in a digest that
was never opened costs more than one that is missing.
Measured against the 1,217 Codex sessions on the machine this was written on: 3,021 of 8,322 commands name a read, 944 sessions gain one, and of 4,472 claimed paths 87.5% are a file that still exists — the rest are the temporary files a session makes and deletes. One was a directory.
A log file it cannot read is named, not skipped. A file whose permissions
changed, or one truncated mid-write into bytes that no longer parse, cannot
contribute a session — and until v0.2.3 it contributed nothing to the report
either, including any mention that it existed. One good file beside two broken
ones reported "1 session", in today, in list and in --json alike, which is
the wrong answer in the direction that looks fine. Every view now ends with
note: 2 session logs are not shown — 1 could not be read, 1 had no readable records
(run with --verbose to see which)
and --verbose names the paths — every one of them, not a sample, because a
report is re-runnable and somebody who asked which files has asked for the list.
It is the same sentence agentwatch prints about the same logs: one wording in
both commands, which arrive in one install. Under --json and --md - the
note goes to stderr, so stdout stays the bare array or the bare document. The exit code
stays 0: agentlog reports, it does not gate. An empty file is deliberately not
one of these — a session that has just started has nothing in it yet, and
nothing has been lost.
Project identification is best-effort. The project is taken from the cwd
field of the first user record in a session. If a session starts before a
user record is written, the project falls back to a guess derived from the
directory name, which is ambiguous when paths contain dashes.
Time nobody was working is not active time. A session is bursts of work with nothing in between, so a stretch of silence longer than five minutes is not counted. Measuring instead from a session's first record to its last — which is what this used to do — billed lunch, the commute and the whole night as work: on the logs this was written against it came to 14× the time Claude Code itself records for the turns in those sessions, and five minutes is the threshold that matches them (0.93 of the recorded time in aggregate, a median of 1.10 per session). Being slightly under is the better way to be wrong for a number somebody might quote.
Each session block therefore prints both — 3m 00s active, 6h 01m open — and
--json gives you active_s beside duration_s, because "how long was this
session open" is a real and separate question.
The headline figure for a period is the union of every session's working stretches clipped to that period, not their sum — agents run in parallel, and summing them produced "111h in a 24h day". It answers "how much of the day had an agent working" rather than "how many agent-hours were spent".
A session that spans the window is counted only for its share of it. A
session running from Tuesday to Friday appears in Wednesday's digest, but only
the files, commands, turns and errors timestamped inside Wednesday are counted,
and the duration shown is what it worked on Wednesday — not the span from
midnight. Leave a session open overnight, come back and run one command at
09:16, and clipping to the edge of the day reported 9h 16m active beside
1 command: every hour spent asleep counted as work.
A session that did nothing at all inside the window is counted as nothing, even
though it still appears there. Only a session whose records carry no usable
timestamps falls back to its lifetime total — that is a session we cannot see
inside, which is a different thing from one we can see slept. Confusing the two
made four separate days each report exactly 24h 00m, and made a week come out
shorter than the days inside it added up to.
Tokens are clipped the same way, against the period you asked for. A
week-long session used to report its whole week's spend into every day it
touched — on one real session, 88.3M printed on the line directly below a
correctly clipped 1 command, against the 14.2M actually spent that day. A
day now counts the turns that spent their tokens inside it, so seven days add up
to their week exactly. Note the window here is the one you asked for, not the
one tightened onto the session's first and last tool call: a reply that costs a
thousand tokens and calls no tool is still spending, and anything spent after the
day's last tool call would otherwise vanish. Sessions whose logs carry no
per-turn record of what was spent still report their lifetime total, for the same
reason a session with no timestamps does.
Consecutive days add up. A day starts at local midnight and ends before
the next one, so anything stamped exactly on the stroke belongs to the day it
opens and is reported by today, not by yesterday. Run the two commands back
to back and each thing that happened is counted once between them.
on DAY names any one of those days: agentlog on 2026-07-31, or on 3d for
three days ago. since cannot answer this — since 2026-07-31 means "from
Friday until now", never "Friday" — so before this the only two days you could
ask about were today and yesterday. on takes a date or a number of days, and
that is the whole of the overlap with since: 10m, 12h and 2w name a
duration rather than a day, so on refuses them rather than rounding, and says
which command wants one. 0d is the one place the two part company on purpose
— since 0d is a window from now until now and is refused as a typo, while
on 0d is today.
Two clocks are involved, and the log's wins. The timestamps come from
whichever machine wrote the log, which may be ahead of the one reading it — an
NTP step, a resumed VM, a synced home directory. A named day runs to midnight,
so events dated a few minutes into the future are still part of today; only
yesterday and an explicit end actually clip. agentlog never treats the
moment you ran it as the end of a window you did not ask to end.
Codex file tracking is partial. Written files come from patch_apply_end
records, falling back to patch envelopes in the command text, which together
cover Codex's normal edit path. A file
changed some other way — sed -i, a heredoc, a script the agent wrote and then
ran — is not detected, and files Codex only reads are never reported. So a
Codex project's file list is a floor, not a complete account.
The digest is a summary, and summaries drop things. Per project it shows
the three most-written files and the three most frequent distinct failures, and
it lists at most eight projects before collapsing the rest into a count.
Failures that differ only below their first line — the same heredoc run three
times — are collapsed into one row with a (3x) marker. A failure the parser
could find no name for is still counted in errors but has no row to print, so
the count can exceed the list. Until v0.2.3 that was most of them: a Codex
patch call carries no command, so every failed patch was nameless. They are now
labelled with the files they touched. --sessions,
agentlog show ID and --json give the unabridged version.
One item is one row. Everything on screen came out of a log file written
by the agent being audited, so a command or path is flattened to a single line
before it is printed. Otherwise a command containing a newline printed as
several rows, each looking exactly like a real command that ran — commands (1): above three $ lines — and agents write multi-line heredocs all day.
The row counts in the headers are the check, and they are only true if this
holds. In agentlog show, a line longer than 400 characters is cut with a
marker saying how much was dropped; --json always has the whole thing.
Tokens are not verified. Token counts come from usage fields in the logs. They may differ from what your billing provider records.
What the agent said is never shown; what you asked for is. agentlog extracts metadata — file paths, shell commands, durations, model names, token counts — plus the prompt you typed that names the work, because a digest that cannot say what a session was for is a list of filenames. Nothing the agent replied, nothing it thought, and nothing a command printed is extracted or displayed by any flag. This is described in full under Privacy below.
No test coverage against the developer's real logs. The test suite uses synthetic fixtures. It cannot guarantee correct parsing of every schema variant in the wild.
Codex parallel workers are merged into one session. When Codex runs worker agents, all of them share one session_id and each writes its own file. They are one session, so they are shown as one row, with the commands and files unioned and the turns, errors and tokens added up. Times therefore span the whole fan -out rather than any single worker, and a per-worker breakdown is not available.
Until v0.2.3 the extra files were discarded and only the richest one kept. On the logs this was found with, that hid 299 of 616 commands in the 21 sessions that had more than one file, and 38 of the 42 discarded files contained commands the kept one did not.
A Claude Code record is counted once, in the file that had it first. Claude
Code writes the same records into two files in two ordinary situations: claude --resume opens a new session id and copies the earlier transcript into it
verbatim before appending anything new, and a copied or moved project directory
leaves the same log under two names, neither a symlink. Each record carries its
own uuid, so a uuid already counted is skipped. Files are read oldest first, by
mtime, which means the sitting where the work actually happened is the one that
reports it — a resume shows the work done in the resume, and the earlier sitting
keeps its own. Timestamps are skipped along with the counts, so a resume does
not inherit the earlier sitting's start time and look like a session that had
been running since the morning.
This is the opposite treatment from Codex above, and deliberately so: Codex's several files are parallel workers whose separate work genuinely adds up, while Claude's are copies of one worker's. Merging is only safe once the records are known to be distinct.
Until this was fixed, on the developer's own 269 Claude sessions the totals were inflated by 163 commands, 14 written files, 20 errors, 101 turns and 31.1M input tokens. No session was lost — a file that is nothing but replay contributes no counts and is not reported as unusable either, since none of its content is missing from the report.
agentlog reads files and prints. No network code. Nothing is uploaded or sent.
One view is the exception, and this is it: the brief. It asks a model to
name what the day's work was, because grouping eleven sessions into "getting the
release out" is a judgement and there is no arithmetic for it. To do that it
hands the day's evidence — project names, the prompts you typed, shell commands,
file paths, the agent's own recaps — to a model CLI already installed on your
machine: claude or codex, preferring whichever one is hosting the shell the
question was asked from. That evidence leaves the machine. Nothing else here does, and
that is not a promise about intentions: agentlog/asking_a_model.py is the only
module in the package allowed to run another program at all, and
tests/test_privacy_claims.py fails if a second one ever does.
The brief is what a bare agentlog prints at a terminal with a model CLI
installed, so at a terminal the evidence goes out by default — the command
says so on stderr while you wait, and --log is the standing way to say no.
Anywhere a person is not looking — a pipe, a hook, a cron job, --json,
--html, --md — the model is never run and nothing is ever sent, whatever
is installed; a script scraping agentlog cannot start a model call it never
asked for. --brief forces the report anywhere, its own help says what it
sends in one line, and no other output mode may be combined with it.
The rest of this section describes every other command.
Half of a conversation is shown, and it is your half. Every mode prints
the prompt you typed that names the work — asked: "fix the parser and run the suite" — because a digest that lists 247 commands and cannot say what any of
them was for is answering a question nobody asks. Somebody wrote down the goal
of the session at the top of it; there is no honest way to report the session
without it.
Not the whole of your half. One prompt per session: the first that says what is wanted rather than agreeing with something ("ok", "yes", "continue" are skipped), cut to 400 characters, and cut again to whatever room the row has.
The agent's half is never shown, and neither is anything a command did:
| shown | not shown |
|---|---|
| the prompt that names the work | anything the agent replied |
| file paths read and written | anything the agent thought (thinking) |
| shell commands, and which failed | anything a command printed (tool_result) |
| durations, model names, token counts | the contents of a file it wrote |
the agent's away_summary recaps |
prompts you queued but never sent |
Two of those rows are less obvious than the rest, so they are worth naming.
A background session ends each turn by writing itself a short recap — "You
asked what's in the ledger: it tracks 104 requests, all done except R102" — as
a system record with subtype away_summary. agentlog shows those. They are
written by the agent about the conversation rather than in it, and Claude Code
had already put the same sentence on your screen.
And two parts of a session file carry message text where nobody looks for it: a
queue-operation record holds the whole of a prompt you typed while the agent
was busy, and a frame-link record holds a question of yours turned into a
heading. There are 4983 of the first and 104 of the second in this machine's
logs. agentlog reads neither — a prompt you queued and closed the terminal on
was never asked, and more than half are never sent (2494 enqueued against 1121
dequeued here); the ones that are get written again as an ordinary prompt when
they go, and that copy is the one agentlog reads.
tests/test_privacy_claims.py is where all of this is enforced rather than
promised: one marker planted in the prompt, which every output mode must show,
and another planted in the agent's replies, its thinking, a tool result, a
written file, a recap's neighbours, a queue-operation and a frame-link,
which no output mode may show — the HTML digest included, since that one leaves
the machine.
The HTML digest may therefore contain file paths, shell commands, recaps, and what you typed to ask for the work. Review it before sharing it with others; the note at the foot of the page says the same thing.
# From source (no install — works immediately):
python3 -m agentlog today
# Install for the current user:
pip install --user .
# Or in a virtualenv:
python3 -m venv .venv
.venv/bin/pip install .
.venv/bin/agentlog todayRequires Python 3.9 or later. Zero runtime dependencies.
agentlog --version prints the version. Worth quoting in a bug report: this
tool reads session files written by somebody else's program, those formats
move, and "which agentlog" is the first question about a digest that came out
wrong.
python3 -m unittest discover -s tests -vThe tests use synthetic JSONL fixtures in a temporary directory. No real home
directory is touched. Pass --home DIR or set AGENTLOG_HOME=DIR to redirect
agentlog to a different home in any context.
Five tools for working with coding agents, same house style: zero dependencies, MIT, no API key, nothing leaves your machine. None of them call a model — that is the point, since the thing being checked already is one.
Each of those four claims is a test rather than a promise, in
tests/test_family_claims.py: every import resolves to the standard library or
to this package, nothing that can open a socket is imported, no environment
variable that looks like a credential is read, and no model SDK or provider
hostname appears anywhere. A claim repeated in five READMEs and checked in none
of them would read as five agreements when it was one assertion.
- stillworks — record what your code does now, catch when it changes later
- agentdiff — see what the agent actually changed, before you merge
- agentlog — what did your coding agent actually do today? ← you are here
- agentwatch — tail what your agent is doing, right now
- unedit — a safety net for letting an agent loose on your files
One install gets all five, and stillworks tools says which ones you have:
pip install stillworks
stillworks toolsMIT. Copyright (c) 2026 stillworks contributors.