fix(opencode): exclude Claude Agent turns by default in focused reports - #2021

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At docs/guide/all-reports.md, line 49:

<comment>This wording implies `ccusage opencode session` matches whole sessions, but focused OpenCode uses session-day matching for every report kind. Distinguish unified session matching from the focused command’s session-day behavior.</comment>

<file context>
@@ -46,7 +46,7 @@ ccusage detects local usage files from Claude Code, Codex, OpenCode, Amp, Droid,
 Unified tables include an **Agent** column so you can compare sources in one view. Focused views remove that comparison layer and show the selected source in more detail where applicable.
 
-When one tool runs another, usage can be logged by both. OpenCode's Claude Agent provider is one example: unified views skip the OpenCode copy of turns that Claude Code's own log has, matching them by session and, in daily views, by day, so a turn that runs across midnight can still be counted twice or partly left out. See [OpenCode](./opencode/#claude-agent-provider).
+When one tool runs another, usage can be logged by both. OpenCode's Claude Agent provider is one example: unified views and `ccusage opencode` skip the OpenCode copy of turns that Claude Code's own log has, matching them by session and, in daily views, by day, so a turn that runs across midnight can still be counted twice or partly left out. See [OpenCode](./opencode/#claude-agent-provider).
 
 ## Supported Sources
</file context>

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At rust/crates/ccusage-adapter-all/src/loader.rs, line 878:

<comment>The date-window match treats a whole OpenCode turn as belonging to its start day, although Claude records usage calls within that turn on their own days. Turns crossing midnight can therefore be counted by both reports or included past `--until`; split or conservatively retain rows that straddle Claude date boundaries.</comment>

<file context>
@@ -853,6 +853,35 @@ impl ClaudeCounted {
+        .map_or(ClaudeCounted::Sessions(HashSet::new()), |(_, counted)| {
+            counted
+        });
+    Ok(without_turns_counted_by_claude(
+        entries,
+        &claude_counted,
</file context>

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At docs/guide/opencode/index.md, line 84:

<comment>The session-report totals do not always add up to `ccusage daily`: Claude session rows retain whole-session usage after selecting sessions by last activity. Limit this claim to daily-family reports and explain the session exception.</comment>

<file context>
@@ -79,7 +79,15 @@ ccusage matches OpenCode rows to Claude usage by Claude Code session, not by ind
-Because OpenCode dates a turn by when it started and Claude Code dates each model call, a turn that runs across midnight can be off in the daily family. It is counted twice when Claude has no usage from that session on the day the turn started. With `--until` on that day, the calls Claude logs after midnight are left out, even though OpenCode dates the whole turn on the earlier day. `ccusage opencode` always shows every `claude-agent` row.
+Because OpenCode dates a turn by when it started and Claude Code dates each model call, a turn that runs across midnight can be off in the daily family. It is counted twice when Claude has no usage from that session on the day the turn started. With `--until` on that day, the calls Claude logs after midnight are left out, even though OpenCode dates the whole turn on the earlier day.
+
+`ccusage opencode` skips the rows that `ccusage claude` counts, matched by Claude Code session and date in every report, session reports included. This way `ccusage claude` and `ccusage opencode` add up to `ccusage daily`, and tools that read each source separately count these turns once. These turns appear under Claude. Pass `--include-claude-agent` to show every `claude-agent` row in the OpenCode report:
+
+```bash
</file context>

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At rust/adapters/opencode/src/README.md, line 19:

<comment>This paragraph now says session reports match whole sessions, then says `ccusage opencode` matches by session day in every report kind. Scope the earlier sentence to unified reports, for example 'unified session reports match whole sessions', so the two rules do not conflict.</comment>

<file context>
@@ -16,7 +16,7 @@ SQLite databases are the primary source. Legacy JSON messages under `storage/mes
 OpenCode v2 forks copy the parent's `session_message` rows with new ids but the same `seq`. When `session_v2.fork_session_id` and `fork_boundary` resolve against the parent's rows, fork rows at or below the last copied `seq` (the boundary for `through`, the parent's last row before it for `before`) are skipped as inherited history. The fork's `session_v2` counters start at zero and hold only fork-local usage, so they remain a valid session fallback. Unresolvable boundaries and pre-`seq` schemas keep every row.
 
-The Claude Agent provider (`claude-agent`) runs Claude Code, which also writes its own transcript. Its v2 rows record that Claude Code session as `content[].native.sessionID`, which `load_entries_with_native_sessions` returns with each entry. Unified daily-family reports drop a row when the Claude source counted main-thread usage of that session on the row's date, in the Claude source's timezone and date range; session reports match whole sessions, since a Claude session row holds all of its usage, using the session ID the transcript records rather than its file name. A resumed or forked transcript repeats earlier turns, sometimes under the original session ID, and Claude's dedup keeps one copy of each. `should_replace_deduped_entry` prefers main-thread copies, then more tokens, then a copy with speed metadata over one without (the daily loader's `should_replace_deduped_daily_entry` compares cost before speed); only among copies equal on all of these does the first read win, with files sorted by path and named by UUID, so which transcript keeps a turn is effectively arbitrary; when all of them are reported, the choice does not affect the total. A session with surviving main-thread usage under its own recorded ID counts as included when the report includes every transcript holding that usage; copies it lost to other sessions are not consulted. A session with no surviving main-thread usage of its own counts as included when every transcript holding its kept copies is included. When a Claude date window includes only some of the transcripts involved, a repeated turn can go either way: a skipped session's rows for it are dropped although its kept copy is in an excluded transcript, or a session keeps its rows when the usage that decides it (its own surviving main-thread usage, or its kept copies when it has none) spans included and excluded transcripts, so the turns Claude counts in the included transcript are counted twice. The usage is then priced by the resolved model ID rather than the `opus`/`sonnet` alias OpenCode stores. The match is per session or session day, not per turn: `native.messageID` is the transcript line `uuid`, and one OpenCode row sums several API calls, so rows do not map one-to-one to Claude entries. This relies on the transcript holding every main-thread turn OpenCode records for that session; it also holds turns OpenCode drops, such as interrupted steps saved with `tokens: null`. A row is dated by its turn's start while Claude dates each call, so in daily-family reports a turn that crosses midnight in the Claude source's timezone is counted by both sources when Claude has no usage of that session on the earlier day, and with `--until` on that day the calls Claude logs after midnight are left out although OpenCode dates the whole turn on the earlier day. Only v2 `session_message` rows can carry the link: OpenCode added the provider after moving to SQLite v2, so legacy JSON messages and the v1 `message` table hold no `claude-agent` rows. `ccusage opencode` keeps every row.
+The Claude Agent provider (`claude-agent`) runs Claude Code, which also writes its own transcript. Its v2 rows record that Claude Code session as `content[].native.sessionID`, which `load_entries_with_native_sessions` returns with each entry. Unified daily-family reports drop a row when the Claude source counted main-thread usage of that session on the row's date, in the Claude source's timezone and date range; session reports match whole sessions, since a Claude session row holds all of its usage, using the session ID the transcript records rather than its file name. A resumed or forked transcript repeats earlier turns, sometimes under the original session ID, and Claude's dedup keeps one copy of each. `should_replace_deduped_entry` prefers main-thread copies, then more tokens, then a copy with speed metadata over one without (the daily loader's `should_replace_deduped_daily_entry` compares cost before speed); only among copies equal on all of these does the first read win, with files sorted by path and named by UUID, so which transcript keeps a turn is effectively arbitrary; when all of them are reported, the choice does not affect the total. A session with surviving main-thread usage under its own recorded ID counts as included when the report includes every transcript holding that usage; copies it lost to other sessions are not consulted. A session with no surviving main-thread usage of its own counts as included when every transcript holding its kept copies is included. When a Claude date window includes only some of the transcripts involved, a repeated turn can go either way: a skipped session's rows for it are dropped although its kept copy is in an excluded transcript, or a session keeps its rows when the usage that decides it (its own surviving main-thread usage, or its kept copies when it has none) spans included and excluded transcripts, so the turns Claude counts in the included transcript are counted twice. The usage is then priced by the resolved model ID rather than the `opus`/`sonnet` alias OpenCode stores. The match is per session or session day, not per turn: `native.messageID` is the transcript line `uuid`, and one OpenCode row sums several API calls, so rows do not map one-to-one to Claude entries. This relies on the transcript holding every main-thread turn OpenCode records for that session; it also holds turns OpenCode drops, such as interrupted steps saved with `tokens: null`. A row is dated by its turn's start while Claude dates each call, so in daily-family reports a turn that crosses midnight in the Claude source's timezone is counted by both sources when Claude has no usage of that session on the earlier day, and with `--until` on that day the calls Claude logs after midnight are left out although OpenCode dates the whole turn on the earlier day. Only v2 `session_message` rows can carry the link: OpenCode added the provider after moving to SQLite v2, so legacy JSON messages and the v1 `message` table hold no `claude-agent` rows. `ccusage opencode` runs through `ccusage_adapter_all::run_opencode`, which matches by session day in every report kind, as `ccusage claude session` counts only in-window days, so focused reports add up to the unified daily one; it reads Claude only when a loaded row carries a native session, treats an unreadable Claude source as counting nothing, and `--include-claude-agent` keeps every row. `opencode::run` keeps every row.
 
 Token mapping:
</file context>

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At rust/crates/ccusage-cli-parser/src/tests.rs, line 1426:

<comment>The test name says it covers every report, but `monthly` is omitted from the loop even though `OPENCODE_AGENT_REPORTS` accepts it. Add `Some("monthly")` so the flag is checked for that report too.</comment>

<file context>
@@ -1418,6 +1418,26 @@ fn parses_opencode_weekly_options() {
+
+#[test]
+fn parses_opencode_include_claude_agent_on_every_report() {
+    for report in [None, Some("daily"), Some("weekly"), Some("session")] {
+        let mut argv = vec!["ccusage", "opencode"];
+        argv.extend(report);
</file context>

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At rust/crates/ccusage-adapter-all/src/tests.rs, line 908:

<comment>This fixture setup is now duplicated. `unified_reports_count_claude_agent_sessions_once` still inlines the same Claude transcript and three OpenCode rows. Have that test call `write_claude_agent_fixture(&fixture)` so the two copies cannot drift apart.</comment>

<file context>
@@ -892,6 +892,130 @@ fn unified_reports_count_claude_agent_sessions_once() {
+        .sum()
+}
+
+fn write_claude_agent_fixture(fixture: &ccusage_test_support::Fixture) {
+    let _ = fixture.write_file(
+        "claude/projects/project-a/claude-native-a.jsonl",
</file context>

Check if this issue is valid — if so, understand the root cause and fix it. When an issue isn't valid or won't be fixed in this PR, reply in its thread with the reason and then resolve the thread. At rust/crates/ccusage-adapter-all/src/loader.rs, line 872:

<comment>This misstates Claude session semantics: session reports select whole sessions by last activity, while this loader matches OpenCode rows by session day. Reword the comment to describe OpenCode's matching intent.</comment>

<file context>
@@ -853,6 +853,35 @@ impl ClaudeCounted {
+        json: true,
+        ..shared.clone()
+    };
+    // Focused Claude reports, sessions included, count only in-window days, so match by
+    // session day. With no readable Claude source, nothing counts these turns and they stay.
+    let claude_counted = load_claude_rows(AgentReportKind::Daily, &claude_shared, None)
</file context>