llm-panel

Several independent LLM code reviews of one diff, on one page, with a measured miss rate.

uv tool install llm-panel            # or: pipx install llm-panel
llm-panel --judges codex --diff "What is wrong with this change?"   # one judge you already have
panel-report --open                  # the run as one HTML page

Any one of codex, claude, opencode or ollama on your PATH is enough to start; the roster is where you add the rest. Unlike a single-reviewer bot, the panel is N readers who cannot see each other, a rebuttal round in which they defend or withdraw, and a recall benchmark that says what they miss.

Judges run in parallel, never see each other's work, and answer from their own reading of your repo. An optional second round shows each of them the others' findings — anonymised — and asks them to defend or withdraw. The output is a single self-contained HTML page where that second round is grouped by the finding being argued about, so comparing what five models said about one line of code doesn't mean holding five documents in your head.

It is not a voting machine. A panel generates candidate defects; it does not establish truth by counting agreements. Every finding still has to be checked against the code — and the tool's other half, recall/, exists to measure what the panel misses rather than assert what it catches.

llm-panel --diff "Which of these changes is most likely to be wrong?"
panel-report --open          # render the newest run and open it
panel-triage --bad           # which runs went wrong, across every run root

A real run, with the waiting compressed: three judges asked in parallel (two free-tier, one on a ChatGPT plan) landing as they finish, the scoreboard from panel.md, and panel-report rendering it (recording):

The rebuttal round as rendered — every position each judge took on each finding, grouped by the finding under dispute, disagreements marked CONTESTED. This run: four free-tier judges asked to review llm-panel's own failure-classification code; one failed and is reported as harness, the other three upheld 7 findings, rejected 4, and missed 6:

Contents: What's here · Beside the review bots · Install · Configure your roster · Using it · What it actually catches · On real PRs · Tests · Known limitations

What's here

tool	what it does
llm-panel	asks the judges, in parallel, and writes the run to disk
panel-report	renders a run as one self-contained HTML page, grouped by claim
panel-triage	finds the runs that failed, which a listing shows as ordinary rows
recall/panel-recall	measures what the panel misses, against a corpus of planted defects
recall/aacr-upstream	runs the panel over AACR-Bench PRs and hands the findings to upstream's evaluator
recall/aacr-score	invokes that evaluator, and refuses to report a number from a judge that isn't running
claimlib.py	the one measurement boundary: reviews → span-grounded observations
*-controls	the regression suites — 1091 controls, every one tied to a defect that shipped

Beside the review bots

	models per review	what you read	published accuracy	runs as	keys
llm-panel	N, independent, blind to each other; rebuttal round	every answer verbatim, grouped by finding	measured on AACR-Bench, low and reproducible	CLI, GitHub Action	yours (subscriptions or API keys)
CodeRabbit	several, in a pipeline by stage	filtered by a verification agent	vendor-reported on Martian's bench	GitHub/GitLab app, CLI, IDE	hosted, no model choice
Qodo PR-Agent (MIT)	one model per call, fallback on failure	structured summary, 3 findings by default	none for the OSS tool	GitHub Action, CLI, Docker	yours
Copilot code review	"a mix of models", not switchable	filtered, severity-labelled	none stated	github.com, gh, IDE	hosted

The difference is not that the panel is better — on the numbers above it is not — but that it shows you everything the models said and tells you how much they miss. A filter that "validates each suggestion" is one more opinion, and the one most likely to drop a minority finding. Verified 2026-09-06 from each vendor's own pages; the smaller open-source council-style reviewers found had under 50 stars, and none publish a miss rate.

Install

Pure Python 3.11+ standard library on Linux, macOS or WSL — it needs POSIX file locks and process groups, and says so on Windows instead of tracing back. No dependencies, no build step. Each tool is one readable file, so either install route runs identical code:

# as a package (entry points: llm-panel, panel-report, panel-triage)
uv tool install llm-panel        # or: pipx install llm-panel

# or as the files themselves
git clone https://github.com/musharna/llm-panel ~/llm-panel
ln -s ~/llm-panel/{llm-panel,panel-report,panel-triage} ~/.local/bin/

3.11 is a hard floor (the link renderer uses atomic groups, added to re in 3.11); panel-report says so at startup rather than failing part-way through a render.

Judges reach models through command-line tools you install separately — none are bundled, and you need at most one to start:

tool	who it is	billing
codex	OpenAI's CLI	a ChatGPT plan, not metered API
opencode	multi-provider CLI most judges route through	your OpenRouter / HuggingFace keys
claude	Anthropic's CLI	a claude.ai subscription (setting ANTHROPIC_API_KEY switches it to metered)
ollama	local models	free, and no tool loop — see the caveat below

A judge whose tool is missing is reported as harness and the panel exits 4 — one judge's problem, never the whole panel's. If no selected judge has its tool, the panel exits 14 and names each one with its install hint.

Two transports skip the CLI: ollama uses its local HTTP API, and orvision calls OpenRouter's HTTP API directly with your OpenRouter key so that the vis-* judges (grok, kimi, gemini, gpt) can look at an --image.

The read-only agent for opencode judges

opencode judges run as an agent named panelist that can read the repo but not write to it, and llm-panel refuses to start an opencode judge until that agent is defined and verified read-only (exit 9) — opencode's default build agent will happily edit the tree it is reviewing. The definition ships as opencode.jsonc: merge its agent.panelist block into ~/.config/opencode/opencode.jsonc, or keep the file at the root of a repo you review with it — opencode reads project-local config too.

Configure your roster

The built-in judge list is a default, not a fixture — it names the author's accounts. Yours will be different. Point the roster at models you actually have:

Copy roster.example.json to ~/.config/llm-panel/roster.json ($XDG_CONFIG_HOME honoured; $LLM_PANEL_CONFIG wins). It is strict JSON — no comments, no trailing commas — and a malformed config is fatal and names the offending key, because quietly falling back to the built-in roster would run a panel you didn't ask for, and bill you for it:

{
  "default": ["codex", "nemotron", "glm", "kimi", "or-deepseek", "or-grok"],
  "judges": {
    "my-gpt": {
      "transport": "opencode",
      "model": "openrouter/openai/gpt-5.6",
      "family": "OpenAI"
    },
    "big-pickle": null
  }
}

null drops a shipped judge. default is the panel run when --judges is absent. llm-panel --list shows the roster offline and marks config-defined judges. llm-panel --check actually pings each one. llm-panel --help-config prints this schema.

Picking judges

"One per vendor" is not the answer. It is tempting to treat vendor labels as a proxy for independent opinions. The evidence says they aren't: Kohli 2026 measured cross-family judge correlation at φ̄=0.389 against same-family 0.437 — barely different — with the three most correlated pairs being cross-family, and found that restricting to one judge per family made effective independence worse (n_eff 1.93 vs 2.18). Family is display metadata here, not policy.

The six-judge set above did score 6/6 against the planted-defect corpus described under What it actually catches, where a two-vendor panel scored 4/6, but treat that as debugging evidence, not as a result: the roster was repaired because of what happened on those very fixtures, so the comparison is in-sample, and the six defects live in only two files (effective n≈2, 95% CI 61–100%).

What to actually do: pick judges by what they find on your code, and use panel-recall to measure it. The quantity worth maximising is each judge's marginal rescue rate — how often it catches something every other judge on the roster missed — not how many logos are represented.

Two practical constraints: wall-clock is the slowest judge, not the sum, so one slow model sets the pace for every run; and claude-* are deliberately absent from the default, because when Claude wrote the code under review a Claude judge shares the author's blind spots. Add it explicitly when that isn't the case — it is strong.

Using it

• --diff attaches the working-tree diff, so nobody has to describe the change — including you, who would describe it favourably.
• --rebut adds the anonymised second round. Worth it whenever a finding would trigger real work: the first run of it killed three confident findings that were simply wrong. To be precise about the report's grouping of that round: it keys on the rebuttal letter each finding is given (A1, B2 …), so it collects the discussion around one judge's finding. It is not semantic clustering — two judges independently raising the same underlying defect stay two findings, and without --rebut there is no grouping at all.
• --judges a,b,c overrides the default panel. codex~2 runs the same model a second time as a full, separate judge — its own file, its own letter, its own row. Collapsing repeats would hide exactly the disagreement that makes them worth running.
• --thread NAME keeps a persistent conversation per judge. For design questions, not review.
• --image PATH (repeatable) attaches an image. Only the vis-* and claude-* judges can look at it; every other judge reports unavailable rather than answering blind, and --vision-check TEXT makes each judge quote something visible before it is believed.
• --live prints each answer the moment it lands instead of waiting for the slowest judge; --stream echoes tokens as they arrive, which only ollama and claude judges can honour. Without either, a heartbeat still names who is still working.
• --runs lists past panels for this repo (--all-repos for every repo) and --show prints the latest report. --effort {low,…,max} sets reasoning effort where a judge has the setting; --timeout SECONDS caps each judge, and a judge over the deadline is killed as a whole process group and reported harness. -f FILE reads the prompt from a file.
• Long questions go via stdin: llm-panel - <<'ASK' … ASK.
• --synthesize JUDGE asks one judge to fold every answer into a single synthesis after the round; --cwd DIR reviews a repository other than the current one; --save-here writes the panel into the reviewed repo as well as the cache; --agent NAME picks the opencode agent (default panelist) and --keep-alive the ollama model residency. panel-report takes --repo SUBSTR to pick a run root, --out FILE, --webfonts and --max-image-kb; panel-triage takes --since HOURS, --repo, --limit and --json.
• --usage shows what the codex judge is spending: the ChatGPT plan's 5-hour and weekly windows, when each resets, and the "Full reset (Weekly + 5 hr)" credits OpenAI banks on the account. --reset-usage redeems one of those credits — it prints the same screen, then asks you to type RESET, because a credit is finite and a script should not be able to spend one by passing a flag. Both read the account through codex app-server, the same channel the interactive /status screen uses, and cost no quota themselves.

On pull requests

The repository doubles as a GitHub Action. It builds a prompt from the PR's diff, runs the panel in the checked-out tree, and posts every judge's answer in full as one comment, edited in place on each push rather than added to:

# .github/workflows/panel.yml
on: pull_request
permissions: { contents: read, pull-requests: write }
jobs:
  panel:
    if: github.event.pull_request.head.repo.full_name == github.repository # forks have no secrets
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v7
      - uses: musharna/llm-panel@v0.1.7
        with:
          openrouter-api-key: ${{ secrets.OPENROUTER_API_KEY }}
          # judges: or-glm,or-kimi,or-deepseek   timeout: "600"   extra-args: --rebut

The default judges are the three OpenRouter ones, so one key is the whole setup. They read the checked-out tree, not just the diff: on this repository's own 9-file PR the three spent 300k–990k input tokens each and billed $0.68 and $1.12 for the panel on two runs, 4.5 minutes wall clock, with kimi-k3 the largest share both times. The job fails on exit 9 — the PR's tree carries .opencode/ or claude hooks the judges would run — and posts the panel on 0 or 4. This repository runs it on its own pull requests (.github/workflows/panel.yml, installing from source).

Judges reading through codex/opencode/claude can read your repo. ollama judges answer from the prompt alone with no tool loop, so they cannot verify a claim against code. Treat their findings accordingly.

Exit codes are deliberate and llm-panel --help lists them: 0 every judge answered · 1 usage, config, or a failure of this program · 2 --file could not be read · 3 --check found a judge that could not answer at all · 4 degraded panel (a judge never ran — our failure, reported as such) · 7/8 --diff could not produce a diff / had nothing to review · 9 the opencode agent or the reviewed tree is not verified safe · 10 --thread is locked by another run · 11/12 --repeat out of range / a repeat suffix typed by hand · 13 illegal judge name · 14 none of the selected judges has its CLI installed, with the roster path in the message · 130 interrupted (Ctrl-C or SIGTERM), with whatever landed kept in the run directory.

The rebuttal round as rendered — every position each judge took on each finding, grouped by the finding under dispute, disagreements marked CONTESTED. This run: four free-tier judges asked to review llm-panel's own failure-classification code; one failed and is reported as harness, the other three upheld 7 findings, rejected 4, and missed 6:

The rendered report — the scoreboard counts spend and names who answered; the citation-overlap tables show where the panel's attention landed (three judges reviewing a cline PR, converging on one line of TerminalProcess.ts):

What it actually catches

recall/panel-recall is the part most tools like this don't have: a corpus of defects planted in real code, each one proven to misbehave by execution, so "the panel missed it" is a measurement rather than an impression.

At least one of four independent passes (codex ×2 + claude-opus ×2) matched 25 of 27 known targets in this controlled, single-file Python corpus. That is a keyword-matched lower bound on an easy corpus — not an estimate of real-world code-review capability. 95% CI 76.6–97.9%, and that is before accounting for defects clustering within fixtures.

Hand-planted single-mechanism defects in ~40-line files are far easier than real defects in mature codebases (CR-Bench reports GPT-5.2 + Reflexion at 32.8% recall on git blamed real bugs), so this corpus is a development instrument for controlled A/Bs, not evidence of capability. Three things it did settle — the two misses were the roster's fault, not the models' (a six-vendor panel found both); running the same model twice recovered nothing where adding a different vendor did; and one sentence of abstention licence trades real findings for zero false positives — are written up with their numbers in recall/README.md.

On real PRs (AACR-Bench)

The real-world numbers come from running the panel over AACR-Bench PRs and scoring the findings with upstream's own evaluator — an LLM judge doing path → line → semantic matching, so the numbers are theirs, not a self-graded matcher's. 18 PRs, full roster, three prompt styles (recall/aacr-upstream --prompt-style). The default row is the shipped prompt measured at e2ad666 (2026-09-06); the other two are re-measurements from 2026-08-28:

--prompt-style	semantic recall	precision	findings read per validated hit
defect (default)	9.8%	9.5%	10.5
broad	26.0%	13.2%	7.6
volume	25.2%	7.9%	12.6

broad doubles defect's recall (McNemar p = 0.0005) for 25% more reading per validated hit; volume reaches the same recall by verbosity alone and halves precision. A declared cost cut kept defect as the default for its precision and names broad the only candidate for a future change (recall/benchmarks/cost-cut/README.md).

Against the paper's own baselines, the panel's precision is ordinary and its recall is low. AACR-Bench's Table 3 (v3, 2026-01-30), single models on all 200 PRs, "No context" — the diff plus title and description, no repository code — is the closest published condition to the diff-in-prompt arm above:

paper, "No context", all 200 PRs	recall	precision
GPT-5.2	47.1%	7.0%
Claude-4.5-Sonnet	42.9%	8.7%
DeepSeek-V3.2	36.5%	5.6%
GLM-4.7	27.6%	11.3%
Qwen-480B-Coder	27.4%	9.4%
this panel, defect, 18 PRs	9.8%	9.5%
this panel, broad, 18 PRs	26.0%	13.2%

The rows are not directly comparable: ours is an 18-PR subsample scored with a different judge model, without the PR title and description, and it is one subscription model plus two free-tier ones against single frontier models. What can be said: defect sits at the low-recall end of that spread, broad sits inside the paper's recall range at better-than-paper precision, and nothing here has been measured on the full 200. The variance floor is measured (effects under ~5–7 pp are re-run noise at this n), three earlier readings were withdrawn on re-measurement and nothing above rests on one, and location agreement overstates semantic agreement ~2.5x — which is why scoring is delegated upstream. The full comparability caveats, every run ledger and the data licensing are in recall/benchmarks/README.md.

Tests

./claimlib-controls              #  90
./llm-panel-controls             # 427
./panel-report-controls          # 319
./panel-triage-controls          #  19
./recall/aacr-upstream-controls  #  96
./recall/aacr-recut-controls     #  27
./privacy-controls               #  10
cd recall && ./panel-recall selftest && python3 validate_corpus.py

CI runs all seven suites on every push (Python 3.11, 3.12 and 3.13).

Every control corresponds to a defect that shipped, and each asserts the fixed behaviour and — where the pre-fix input is representable — that the broken version would have failed on it. An assertion that passes on both the broken and the fixed code tells you nothing.

Known limitations

• The judge roster's shipped defaults will not work for you until you configure it.
• Judges can read the working tree. --diff sends untracked file contents to remote APIs. Don't point it at a repo holding secrets you haven't gitignored.
• Reviewing a repository means trusting its .opencode/, opencode.json[c], .claude/settings*.json hooks and .mcp.json. opencode loads plugins, tools and agent definitions from the tree it is pointed at, so a repository can ship code that a judge would run as you. llm-panel refuses (exit 9) when the tree carries any of that. claude -p was measured to run a tree's .claude/settings.json hooks with no trust prompt, so a claude judge is refused the same way when the tree declares hooks or .mcp.json; --unsafe-agent overrides all of it.
• A prompt over 128 KB is written to <repo>/.llm-panel-material/ for the run so judges' read tools can reach it; it is removed when the run ends, on any exit. On a shared host, the prompt is also visible in the judge processes' command lines while they run.
• A panel is not a jury. Independent models generate candidates; verification against code, tests, and execution is still yours to do.
• Recall is measured on a 27-defect Python corpus. That number does not transfer to other languages or to defect classes the corpus doesn't contain.
• The headline recall numbers are prompt- and condition-specific. They move with --prompt-style, roster health, and diff-vs-checkout context — see On real PRs before quoting any of them.
• Every other fixture has verified scope, not proven absence. Their known unplanted defects are recorded in each truth.json and re-checked by execution in validate_corpus.py, so a judge that finds one is not scored as wrong. Anything not yet recorded still depresses the recall floor by making a true finding look like noise.

License

MIT — see LICENSE. The MIT grant covers the code in this repository; the benchmark data under recall/benchmarks/ contains third-party material (PR diffs and review-comment text) that stays under its upstream terms — see recall/benchmarks/PROVENANCE.md.