Skip to content

Make syberlabs experiment report read the GitHits pilot's API arm from the host's per-call launcher ledger (signal.githits), so arm-B items are compared instead of excluded as arm_conflict. - #141

Merged
3 commits merged into
mainfrom
fix/experiment-report-githits-api-arm
Oct 27, 2026

Conversation

@sdcarlson

@sdcarlson sdcarlson commented Oct 26, 2026 •

Copy link
Copy Markdown
Contributor

Summary

Since #134, syberlabs experiment report excluded every arm-B item of an API-route GitHits pilot as arm_conflict, so it compared nothing and could give no go/no-go. It now reads arm B from the host's per-call launcher ledger (signal.githits), and the pilot's go/no-go comes out decided. Three commits:

  1. be67706: the fix in syberlabs/experiment.py.
    • It reads arm B from signal.githits.
    • It never reads an unknown lookup count as zero: use_unknown (new, kept) and contamination_unknown.
    • It adds tests, the rehearsal's --route api, and docs/GITHITS.md.
  2. 86e509b: two_routes (new, kept), for a stray GitHits MCP server beside the launcher in arm B.
  3. d10fc3d: docs/EXPERIMENTS.md and the --server help.
    • The doc covers both routes, the protocol commands for the API route, and what to do on use_unknown and two_routes.
    • The --server help no longer implies MCP is the only route.

What was wrong

PRs #129 and #134 moved arm B of the GitHits pilot from an MCP server to the host's API client: ClaudeCodeProvider(githits=True) writes a per-call launcher, delivers the adapted skill, and journals one lookup fact per authentic client record. #134 taught syberlabs/metrics.py to count those lookups, but syberlabs/experiment.py still told the arms apart only by an MCP server: SERVER = "githits", read from signal.harness.mcp_servers (githits:connected). #135 did not touch experiment.py.

I proved it end to end before changing anything, and again on the current main (4219850, after #135):

  • Both arms ran through the real ClaudeCodeProvider: arm A with githits=False, arm B with githits=True and one launcher search.
  • The run used the stand-in claude from tests/test_claude_code.py, the loopback FakeGitHits, and the real query guard and launcher.
  • Each thread was labelled with syberlabs experiment label before its first candidate, then accepted.

Results on main:

  • Arm B's candidate had signal.githits = {"arm": "on", "launcher_calls": 1, "launcher_passed": 1, "records": 1, ...}, no mcp_servers in signal.harness, and one githits.search lookup fact.
  • experiment.runs read server as False for both arms.
  • experiment.report returned items {'labelled': 1, 'compared': 0, 'excluded': 1}, with flags ['arm_conflict'] and excluded ['arm_conflict']. The rule was met: None ("no item delivered in both arms").
  • syberlabs experiment report --name githits-pilot exited 1. It printed check arm_conflict (excluded): LIB-1: the record contradicts the label: the server was connected in the baseline, or not in the treatment, then no item ran in both arms without a check that excludes it, so nothing was compared.

So every arm-B item of an API-route pilot is excluded, and the go/no-go cannot be computed. --server none would hide the problem only by dropping the arm, contamination and use checks altogether. On this branch the same run gives compared: 1, no check, server True for arm B, and exit 0.

The fix

How the report identifies the arm. Per call, the report now counts GitHits as present when the candidate carries signal.githits (for --server githits, the default), as well as in the MCP form. Why this field, after reading claude_code.py:

What each check now means (existing names unchanged; two kept checks added, use_unknown and two_routes):

  • arm_conflict (excluded): a baseline call had GitHits (the ledger, or the server connected), or a treatment thread's calls had none.
  • contaminated (excluded): the baseline shows a GitHits lookup. That is either its githits_lookups is above zero, or its thread holds a lookup fact the host observed and can vouch for: an mcp__githits__* call, or a githits.<op> client record that a guarded launcher call stands behind (guard pass). The second reading also catches a lookup from a call that failed before its candidate: such a lookup is journaled but never counted by a registration. As in metrics since Harden the GitHits launcher route before the A/B pilot: the launcher entry accepts only the five lookup operations with license_mode strict, arm A loses every GitHits variable, the skill digest is LF- #135, a githits.doctor fact is a readiness check, not a lookup.
  • contamination_unknown (kept, as before): the baseline shows no lookup and its count is unknown.
  • unused (kept): the treatment shows no lookup and its count is known to be zero. A launcher call the guard refused never reached GitHits.
  • use_unknown (new, kept): the treatment shows no lookup and its count is unknown, so it can be shown neither to have used GitHits nor not to.
  • server_mixed (kept): a treatment thread had GitHits in some calls only. A baseline thread like that is arm_conflict.
  • two_routes (new, kept; follow-up commit): a treatment thread had GitHits by both routes: the ledger, and a connected GitHits MCP server left in arm B by mistake.
    • It is still the treatment: GitHits was there, so it is not arm_conflict.
    • It is said because it is not the single route the pilot runs. Also, where client records exist, metrics counts only those and leaves the MCP lookups out (spec/GAPS.md 59, item 14).
    • It is kept, not excluded, because the pilot pre-registers its exclusions.
    • A stray server in arm A stays arm_conflict, and its lookups contaminated.
  • arm_unverified (kept): no call recorded its MCP servers or the ledger.

MCP-era records are read as before: a call counts by either route, and --server <other> never reads the launcher. tests/test_experiment.py is unchanged and passes.

How this interacts with #139. With #139's metrics, githits_lookups becomes unknown when the journal cannot say how many lookups reached GitHits: a call's signal.githits.mismatch, or an unguarded client record. The report never reads that unknown as zero:

With main's metrics the same journals read unused and contaminated (it counts what it has). The new tests assert whichever outcome matches the count metrics returns, so they assume neither reading. I checked them against #139's commit cherry-picked onto this branch in a scratch worktree, with its metrics.py hunk merged with #135's doctor exclusion: my tests and tests/test_experiment.py pass (47 tests). There, a mismatch treatment reads use_unknown and an unguarded baseline record reads contamination_unknown. (#139's own test_githits_accounting and one test_metrics assertion need adjusting for #135 when #139 rebases. That is unrelated to this change.)

Rehearsal. examples/githits_ab.py gains --route api, now the default, alongside --route mcp (the previous behaviour):

  • On the API route both arms run the real ClaudeCodeProvider on the api lane, arm B with githits=True.
  • The stand-in claude is self-contained: it prints stream-json, runs the --settings hook before each tool call, and its Bash tool runs only the launcher. GitHits is a loopback stand-in.
  • LIB-4's treatment runs without --githits and is excluded.

Docs. I updated the experiment module docstring, and added a section "The A/B report" to docs/GITHITS.md, placed after "The pinned contract". docs/EXPERIMENTS.md (third commit) now covers both routes:

  • "How the arm is read";
  • the pilot protocol's commands on the API route, with a note on the MCP era;
  • the rewritten check rows;
  • two_routes and use_unknown, with what an operator should do.

The --server help in syberlabs/cli.py now says that for githits the host's API launcher counts too. In docs/GITHITS.md: It covers the commands for both arms, what the report reads, the checks on this route, unknown counts, and the limit: a proposer that writes its own signal could declare a ledger, just as it could declare an MCP server. The doc change is additive only, and #135's text is unchanged.

Evidence

  • New tests. tests/test_experiment_githits_api.py adds 12 tests:

    • synthetic API-route checks: clean pair, conflict both ways, contamination by host-observed facts, unused, unknown counts, mixed, ledger shape, both routes, other servers;
    • an end-to-end run of both arms through ClaudeCodeProvider, labelled and reported with the CLI;
    • arm B with a stray GitHits MCP server beside the launcher (githits=True plus --mcp-config), and the same server in arm A;
    • a misconfigured API pilot;
    • the rehearsal.

    Against main's experiment.py they fail (15 failures and 1 error). There the end-to-end reads compared: 0, excluded: 2 and the API-route rehearsal reads compared: 0, excluded: 7.

  • Printed result. The end-to-end test prints the paired result and the rule's verdict:

    GitHits pilot, API route, both arms through ClaudeCodeProvider (stand-in claude, loopback GitHits):
      LIB-1     off  cost $0.1518  turns 4  fix rounds 1  web 2  githits 0  reviewed 1
      LIB-1     on   cost $0.0759  turns 2  fix rounds 0  web 0  githits 1  reviewed 1
      ROUTINE-1 off  cost $0.0759  turns 2  fix rounds 0  web 0  githits 0  reviewed 1
      ROUTINE-1 on   cost $0.0759  turns 2  fix rounds 0  web 0  githits 1  reviewed 1
      rule (accepted and reviewed = delivered): met. Delivered 2 on vs 2 off; median paired cost over 2 items delivered in both: -$0.0379
    
  • Rehearsal, API route. python examples/githits_ab.py: 6 items compared, check arm_conflict (excluded): LIB-4 and check unused (kept): LIB-4. The rule is not met overall and met on needs_lookup, as docs/EXPERIMENTS.md says. The loopback GitHits received 9 requests. --route mcp gives the same items, checks and verdicts as before.

  • Full suite. env -u ANTHROPIC_BASE_URL -u CLAUDE_CODE_PROVIDER_MANAGED_BY_HOST PYTHONPATH=. python3 -m unittest discover -s tests on this branch, rebased on 4219850: Ran 1007 tests in 828.085s, OK (skipped=24) at be67706; Ran 1008 tests in 744.475s, OK (skipped=24) at 86e509b; Ran 1008 tests in 828.271s, OK (skipped=24) at d10fc3d. Conformance was re-run at d10fc3d and matched 23 golden traces.

  • Conformance. PYTHONPATH=. python3 -m conformance.run: matched 23 golden traces.

  • CI at d10fc3d: every check is green in both the pull_request run and the push run: Python 3.11–3.13, Windows, compose, G2–G8, admit, policy.

    • The push run's first Windows attempt failed in the sharded unit-test step. I couldn't fetch its log here: the session proxy blocks the blob-storage log URL.
    • The same commit passed Windows in the pull_request run, and again when the failed job was re-run. The commit changes only a doc and a help string, which no test reads, so it was a flake.
  • Critic. The reference critic's integrity rules (examples/critic_rules.py) find nothing in the three changed Python files.

The numbers come from stand-ins: they exercise the pipeline and say nothing about GitHits.

Rehearsal: a valid go/no-go on the API route

python examples/githits_ab.py (default --route api), run at 86e509b, sends seven items through both arms of the real ClaudeCodeProvider, with a stand-in claude and a loopback GitHits. Simulated: the numbers are invented to exercise the pipeline and say nothing about GitHits. The go/no-go comes out decided:

  • overall (--delivered review): not met, because the median paired cost over the 5 items delivered in both arms is +$0.0027;
  • on needs_lookup: met, at $0.0321 -> $0.0160 per item;
  • LIB-4, whose treatment ran without --githits, is excluded as arm_conflict, with 6 of 7 items compared.

On main the same rehearsal compares nothing (compared: 0, excluded: 7).

Full output of python examples/githits_ab.py
SIMULATED: a stand-in answers instead of a model. These numbers check the pipeline; they are not results.
Arm B's GitHits route: api

LIB-1      off  2 candidate(s), acceptance succeeded, review pass
LIB-1      on   1 candidate(s), acceptance succeeded, review pass
LIB-2      off  2 candidate(s), acceptance succeeded, review fail
LIB-2      on   1 candidate(s), acceptance succeeded, review pass
LIB-3      off  2 candidate(s), acceptance succeeded, review pass
LIB-3      on   1 candidate(s), acceptance succeeded, review pass
LIB-4      off  2 candidate(s), acceptance succeeded, review pass
LIB-4      on   2 candidate(s), acceptance succeeded, review pass
ROUTINE-1  off  1 candidate(s), acceptance succeeded, review pass
ROUTINE-1  on   1 candidate(s), acceptance succeeded, review pass
ROUTINE-2  off  1 candidate(s), acceptance succeeded, review pass
ROUTINE-2  on   1 candidate(s), acceptance succeeded, review pass
ROUTINE-3  off  1 candidate(s), acceptance succeeded, review pass
ROUTINE-3  on   1 candidate(s), acceptance succeeded, review pass
The loopback GitHits received 9 requests; the real API received none.

Per work item (syberlabs metrics):
thread    class  started           cost                          turns  hours    fix  web  githits  opened  delivered  rejections  interventions
057e66d0  R      2026-10-26 21:42  $0.0134 (host_observed, api)  2      unknown  0    0    1        no      no         0           0
1076b935  R      2026-10-26 21:41  $0.0321 (host_observed, api)  4      unknown  1    4    0        no      no         0           0
18ec42c0  R      2026-10-26 21:42  $0.0106 (host_observed, api)  1      unknown  0    0    0        no      no         0           0
1cd8f2b8  R      2026-10-26 21:42  $0.0106 (host_observed, api)  1      unknown  0    0    0        no      no         0           0
42057879  R      2026-10-26 21:41  $0.0321 (host_observed, api)  4      unknown  1    4    0        no      no         1           0
433ecc7b  R      2026-10-26 21:41  $0.0160 (host_observed, api)  2      unknown  0    0    2        no      no         0           0
6b7f6b8c  R      2026-10-26 21:42  $0.0134 (host_observed, api)  2      unknown  0    0    1        no      no         0           0
7f32b128  R      2026-10-26 21:42  $0.0134 (host_observed, api)  2      unknown  0    0    1        no      no         0           0
a22e42b0  R      2026-10-26 21:41  $0.0321 (host_observed, api)  4      unknown  1    4    0        no      no         0           0
a5ce192c  R      2026-10-26 21:41  $0.0160 (host_observed, api)  2      unknown  0    0    2        no      no         0           0
bba636c6  R      2026-10-26 21:42  $0.0321 (host_observed, api)  4      unknown  1    4    0        no      no         0           0
c8ba42b8  R      2026-10-26 21:41  $0.0160 (host_observed, api)  2      unknown  0    0    2        no      no         0           0
d2543d0f  R      2026-10-26 21:41  $0.0321 (host_observed, api)  4      unknown  1    4    0        no      no         0           0
d507af13  R      2026-10-26 21:42  $0.0106 (host_observed, api)  1      unknown  0    0    0        no      no         0           0

Arms compared (syberlabs experiment report --delivered review):
experiment githits-rehearsal: on against off, 6 items compared, 1 excluded, 14 threads
metric               pairs  off mean  on mean  difference  change  w/l/t  p     95% interval
cost                 6      $0.0214   $0.0147  -$0.0067    -31%    3/3/0  1     [-$0.0175, $0.0041]
turns                6      2.5       2        -0.5        -20%    3/3/0  1     [-2.22, 1.22]
time_to_pass         6      0.0002    0.0002   0           +0%     3/3/0  1     [-0.000115, 0.000115]
fix_rounds           6      0.5       0        -0.5        -100%   3/0/3  0.25  [-1.07, 0.0749]
web_lookups          6      2         0        -2          -100%   3/0/3  0.25  [-4.3, 0.3]
githits_lookups      6      0         1.5      1.5         -       -      -     [0.925, 2.07]
review_rejections    6      0         0        0           -       0/0/6  -     -
human_interventions  6      0         0        0           -       0/0/6  -     -
accepted             6      1         1        -           -       0/0/-  -     -
reviewed             6      0.833     1        -           -       1/0/-  1     -
rule (accepted and reviewed = delivered): not met. Delivered 6 on vs 5 off; median paired cost over 5 items delivered in both: $0.0027
by tag needs_lookup (3 items): cost $0.0321 -> $0.0160 (-50%), turns 4 -> 2 (-50%), fix_rounds 1 -> 0 (-100%), web_lookups 4 -> 0 (-100%), githits_lookups 0 -> 2; rule met
by tag routine (3 items): cost $0.0106 -> $0.0134 (+25%), turns 1 -> 2 (+100%), fix_rounds 0 -> 0, web_lookups 0 -> 0, githits_lookups 0 -> 1; rule not met
  off: 6 threads, 6 accepted, 0 pull requests opened, 0 merged, agent time to merge known for 0, 0 lookups blocked by the guard
   on: 6 threads, 6 accepted, 0 pull requests opened, 0 merged, agent time to merge known for 0, 0 lookups blocked by the guard
check arm_conflict (excluded): LIB-4: the record contradicts the label: the server was there in the baseline (connected, or for GitHits the host's API launcher), or not in the treatment
check unused (kept): LIB-4: the treatment made no lookups on its server
forgotten threads: not visible in exported records
w/l/t: items where the treatment was better, worse, equal. Efficiency compares items accepted in both arms. p: exact sign test. Unknown is never zero.

As the write-up prints it (--markdown):
| Metric | Unit | off (mean) | on (mean) | Change | Items (w/l/t) | p |
| --- | --- | --- | --- | --- | --- | --- |
| Cost | $ per item | $0.0214 | $0.0147 | -31% | 6 (3/3/0) | 1 |
| Turns | turns per item | 2.5 | 2 | -20% | 6 (3/3/0) | 1 |
| Time to a passing change | hours per item | 0.0002 | 0.0002 | +0% | 6 (3/3/0) | 1 |
| Fix rounds | retries per item | 0.5 | 0 | -100% | 6 (3/0/3) | 0.25 |
| Web lookups | searches per item | 2 | 0 | -100% | 6 (3/0/3) | 0.25 |
| GitHits lookups | calls per item | 0 | 1.5 | - | 6 | - |
| Review rejections | rejections per item | 0 | 0 | - | 6 (0/0/6) | - |
| Human interventions | interventions per item | 0 | 0 | - | 6 (0/0/6) | - |
| Accepted | share of items accepted | 1 | 1 | | 6 | - |
| Accepted and passed review | share of items accepted and passed review | 0.833 | 1 | | 6 | 1 |

| Tag | Items | Cost | Turns | Fix rounds | Web lookups | GitHits lookups | Rule |
| --- | --- | --- | --- | --- | --- | --- | --- |
| needs_lookup | 3 | $0.0321 -> $0.0160 (-50%) | 4 -> 2 (-50%) | 1 -> 0 (-100%) | 4 -> 0 (-100%) | 0 -> 2 | met |
| routine | 3 | $0.0106 -> $0.0134 (+25%) | 1 -> 2 (+100%) | 0 -> 0 | 0 -> 0 | 0 -> 1 | not met |

Rule (accepted and reviewed = delivered): not met. Delivered 6 on vs 5 off; median paired cost over 5 items delivered in both: $0.0027.

6 items compared, 1 excluded. Checks: arm_conflict (excluded): LIB-4; unused (kept): LIB-4. Forgotten threads: not visible in exported records. Method: item is the unit; finishing over every valid item, efficiency over items accepted in both arms; pairs where both arms' value is known; unknown counted, never imputed; repeats averaged, unequal repeats excluded; every paired number counted up to the first acceptance; exact two-sided sign test; 95% t interval of the mean difference; no correction for several metrics.

Related open pull requests

Left to others

An agent session built this directly in a worktree, not through syberlabs propose.

Class: R
Agent-platform: claude
Admitted-by: self (Class R)

Claim ceiling: implemented and tested (stand-in claude, loopback GitHits). Not measured; no pilot has run on it.

https://claude.ai/code/session_01HCTqqTVthqCzZJu6MWasuz


Generated by Claude Code

…he experiment report

PRs #129 and #134 moved arm B of the GitHits pilot from an MCP server to the host's API
client (ClaudeCodeProvider(githits=True)), but syberlabs/experiment.py still told the arms
apart only by signal.harness.mcp_servers listing githits:connected. Every arm-B thread of an
API-route pilot therefore read "not connected" and every item was excluded as arm_conflict:
run end to end through the real provider (stand-in claude, loopback GitHits, real guard and
launcher), the report compared 0 items and the CLI exited 1 with "nothing was compared".

The report now also counts a call as having GitHits when its candidate carries
signal.githits, the ledger ClaudeCodeProvider writes for every call it gave the launcher, and
for no other, holding the host's own count of the client records it verified (records). That
is the strongest host-owned per-call evidence the provider records: written only after the
host wrote the call's launcher and record directory and read them back, and present whether
or not the model used it. Model-driven counts (signal.tool_usage, launcher_calls), the skill
text (signal.harness.skill) and the operator's arm name (signal.githits.arm) are not read for
the arm. The MCP route is read as before; existing tests are unchanged.

contaminated and unused also read the lookup facts the host observed (an mcp__githits__* call,
or a githits.<op> client record a guarded launcher call stands behind; never githits.doctor),
so a lookup from a call that failed before its candidate is seen. An unknown githits_lookups
(a call that declared no count, or, with #139's metrics, signal.githits.mismatch or an
unguarded record) is never read as zero: a baseline that shows no lookup is then
contamination_unknown, and a treatment use_unknown (new, kept), never unused. arm_conflict,
server_mixed and arm_unverified keep their meaning per call.

tests/test_experiment_githits_api.py adds synthetic checks for the API route, an end-to-end
run of both arms through ClaudeCodeProvider labelled and reported with the CLI (it prints the
paired result and the rule's verdict), a misconfigured pilot, and the rehearsal.
examples/githits_ab.py gains --route api (default: the real provider with its own stand-in
claude and loopback GitHits) beside --route mcp. docs/GITHITS.md documents the report on this
route.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HCTqqTVthqCzZJu6MWasuz
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Oct 26, 2026 •

Copy link
Copy Markdown

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-10-26T21:40:21.260632Z be67706 PR opened
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: be677064a2

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment thread syberlabs/experiment.py
Comment on lines +275 to +276
if tool.startswith(f"mcp__{server}__"):
found += 1

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Exclude guard-blocked MCP attempts from observed lookups

When an MCP lookup is blocked by the query guard and its provider call fails before registering a candidate, this branch still counts the attempt as evidence that GitHits was used. Transcript.lookups() in syberlabs/claude_code.py journals blocked attempts too, with the blocking rule in guard and ok: false; a verified observation therefore does not establish that the request reached the server. This newly excludes an otherwise clean baseline as contaminated, or suppresses unused for an idle treatment, even though the guard prevented the lookup. Check the guard outcome before counting MCP facts as observed use, as the client-record branch already does.

Useful? React with 👍 / 👎.

sdcarlson and others added 2 commits October 26, 2026 14:42
…tHits MCP server

A treatment thread whose calls had both the host's API launcher (signal.githits) and a
connected GitHits MCP server is still the treatment, so it is not arm_conflict, but it is not
the single route the pilot runs. Where it has client records, metrics also counts only those
and leaves the MCP lookups out (spec/GAPS.md 59). The report now says so with a kept check,
two_routes. It is not excluded, because the pilot pre-registers its exclusions. A stray server
in arm A stays arm_conflict, and its lookups contaminated.

tests/test_experiment_githits_api.py runs it end to end through ClaudeCodeProvider with
githits=True and --mcp-config, plus the same stray server in arm A. docs/GITHITS.md lists the
check.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HCTqqTVthqCzZJu6MWasuz
…ent report's --server help

docs/EXPERIMENTS.md described only the MCP route. It now covers both routes. On the API
route (--githits, since #134), arm B is identified from signal.githits, the host's per-call
launcher ledger; on the MCP route (--mcp-config), from signal.harness.mcp_servers. The doc
also says how a GitHits lookup is shown and that an unknown count is never read as zero.

The pilot protocol's commands move to the API route, with a note on the MCP era. The
checks table rewrites the arm_conflict, contaminated, arm_unverified, server_mixed, unused
and contamination_unknown rows, and adds two_routes and use_unknown with what each means and
what an operator should do.

The --server help in syberlabs/cli.py no longer implies an MCP server is the only route: for
githits it also covers the host's API launcher. The edit is to a line that the open #137
and #138 do not touch.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HCTqqTVthqCzZJu6MWasuz
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant