agents-tree vs BotGauge

agents-tree 60

Self-hosted rule. No payment protocol (0). Free under MIT with nothing to buy (20). No card or account (20). A person installs it from PyPI with one command and no sign-up (20). 60 in all.

BotGauge 10

No x402 or other machine payment (0). Paid plans for the agent product have no published prices, and botgauge.com/pricing prices the separate QA service (0). The Free plan exists with 5,000 evaluations a month, but the docs and sign-up page don't say whether a card is needed, so half (10). A person signs up in a browser at cloud.botgauge.com and creates keys in the console (0). 0+0+10+0 = 10.

agents-tree 60

Local-package reading, as for agent-tree, superpowers and tokeneconomizer. Installs from PyPI with uv tool install agents-tree or pipx, with Python 3.10 or later and POSIX stated in pyproject.toml and Linux and macOS in the herdr manifest (20). .github/workflows/test.yaml runs pylint, pyright, 173 pytest tests, a test run and a package build on Python 3.10 to 3.14 on every push and pull request. Whether the runs pass wasn't checked, as github.com couldn't be read (15 of 25). Issues couldn't be read for the same reason (5 of 25, unchecked). Semver tags and Conventional Commit messages, but no changelog, and the README's caveat that transcript formats may change is the only note on breakage; the switch to reading every provider by default came with 1.0.0 (5 of 15). Version 1.2.0 (15). 60 in all.

BotGauge 18

No status page. status.botgauge.com doesn't resolve and neither the site nor the docs link one (0). With no history to read, 5. No request rate limits for ingest or the eval API. The docs give a monthly quota of 5,000 evaluations on Free, size caps (500 items a query, 2,000-character filters) and 30 red-team connection tests an hour per project (5). No 429 or Retry-After guidance. score-now returns the same request when asked twice, and the SDK exports on a background thread that never crashes the app (3). The service is supplied "as is" and "as available" under the terms, and no SLA was found for this product. The 99.9 per cent SLA on the /ai-agents page belongs to the QA service (0). The service isn't labelled beta, but the SDK is 0.x, marked "4 - Beta" on PyPI, and the docs say API shapes may change before 1.0 (5). 0+5+5+3+0+5 = 18.

agents-tree 73

Framework reading, how much an agent has to carry. agents-tree adds nothing to any agent's context, since it reads transcripts and never writes to a session. Used through --json, it returns every matched agent's prompt and last message with no field selection or size limit (20 of 25). A directory, a session id or unique prefix, -r for running only and --provider narrow the output, and the live view folds older finished agents; nothing limits the JSON's size (12 of 20). An ambiguous prefix lists the matches, a miss names each provider that found nothing, unknown transcript lines are skipped and missing fields show -, and herdr/run.sh names the Python it needs (14 of 20). Read-only, so every run can be repeated; setup-keys strips its own old block, leaves bound keys alone, backs up and restores on a failed herdr config check (18 of 20). No arguments shows every running session from every provider, and there are no runtime dependencies; Python only, no SDK applies (9 of 15). 73 in all.

BotGauge 55

API responses can be narrowed by subject_ids and subject_kind, a results query takes up to 500 items, and judges read at most 20,000 characters, but no field selection or summaries are documented (12). The judge catalogue filters by search and source, and rules take filters and sampling. No pagination is documented on any list endpoint (10). Errors carry a code, a message and a request id. Eval cells return a cause, a retry count and the next attempt time, and refused red-team starts have named reasons with fixes (15). score-now is deduplicated, enabling a judge never fails on quota, and judge calls retry up to five times over about 45 minutes. No idempotency keys on other writes (8). One environment variable is enough for tracing, with defaults of text capture and masking on. Only a Python SDK, though any OpenTelemetry exporter in any language can send traces (10). 12+10+15+8+10 = 55.

agents-tree 64

Framework reading, docs a model can follow, as agent-tree was read. --json writes one document with schema_version 1 and a fixed set of session, agent and detail fields, and model.py types them as dataclasses, but no JSON Schema or field reference is published (14 of 25). README, AGENTS.md and docs/demo are Markdown in the repository; no llms.txt (10). The README says what it shows and reads, has a caveats section per provider (context window inference, internal formats, how running sessions are found), and the command's help says not to run it inside an agent (15 of 20). argparse with choices for --provider and --color, positive-number checks for --window and --interval, and --json refused with --watch (10 of 15). Usage examples for every flag, key and column tables, and errors as messages on stderr with exit code 1, with few error outputs documented (10 of 15). Semver tags and a schema version, no changelog (5 of 15). 64 in all.

BotGauge 53

No OpenAPI file for the eval or red-team endpoints was found in the docs or llms-full.txt. Ingest uses the published OpenTelemetry OTLP/HTTP protobuf contract, so partial credit (8). llms.txt, llms-full.txt and Markdown for every page on Mintlify (10). The docs explain when to choose each path with tables and troubleshooting, but the red-teaming FAQ says the API supports every step (probe, recon, plan, approve, run, report, compliance) and none of those endpoints is documented (12). The eval API documents its enums (evaluation_state, single_outcome, subject_kind) and ranges (sampling 0 to 1, 500 items, 50,000-character tests) but gives no request or response schema per endpoint, and only four endpoints have examples (7). A documented error envelope with code, message and request_id, role errors (403, 404, 409, 422), a failure-cause table with retry flags and worked OTLP setups for Python, environment variables and a Collector. No list of error codes (11). /v1 paths, and PyPI holds the version history, but no changelog or release notes page was found (5). 8+10+12+7+11+5 = 53.

agents-tree 66

Harness reading of the framework checklist, as agent-tree and superpowers were read. 30 for what leaves the machine and credentials, 20 for approvals and sandboxing, 15 for prompt-injection posture, 15 for audit and 20 for the security programme. No credentials, no network code and no runtime dependencies; PyPI releases are built by trusted publishing with no stored token and wait for a maintainer to approve the pypi environment. It reads every transcript of four agent CLIs, prompts included, and the herdr plugin runs the repository's bundled source rather than a PyPI release (26). Read-only by design. SQLite files are opened with mode=ro, nothing is written to an agent's files, and the only writes are a pane mark (directory 0700, file 0600, O_NOFOLLOW) and herdr's config block with a backup and check. It can't act on an agent (17). Control characters and escape sequences are stripped from every transcript field before a view, and the help warns that output run inside an agent lands in its context, but --json passes prompts and last messages unredacted with no note that they are untrusted (10). It keeps no state or log of its own. The detail view names each transcript's path, and herdr's config keeps a backup (8). No SECURITY.md, no security.txt and no advisories found; GitHub's security page couldn't be read. CI pins third-party actions to commit SHAs with Dependabot on them (5). 66 in all.

BotGauge 56

Project API keys reach one project, are shown once, stored in a form that can't be read back per the DPA, show their last-used time and revoke immediately, with documented rotation steps. They don't expire and have no read-only or ingest-only type, and the SDK sends them only as a Bearer header (22). Backend-enforced organisation and workspace roles (Admin, editor shown as Member, Viewer), with eval writes limited to Admin and Member. Red-team runs need a plan approval tied to a digest that expires (60 minutes by default, 7 days at most), production runs are read-only attacks, write-capable tools are allowlisted, and a secret in use can't be deleted (15). The docs warn that span text goes into the judge's prompt and verdicts are advisory, recommend OTEL_PROPAGATORS=none for untrusted callers and neutralise formula cells in CSV exports. No guidance for agents reading traces or transcripts back (8). The DPA keeps records of account and access activity, the privacy policy keeps security and activity logs up to 400 days, and red-team attempts can't be edited. No customer-facing audit log is documented (5). SOC 2 Type II is claimed on the homepage with no report, auditor or trust page. No security.txt, disclosure policy or bug bounty found. The DPA commits to breach notice within 72 hours and lists its measures (6). 22+15+8+5+6 = 56.