Can a cheap model install and use TayHCode/tailscale-agent-access?

Newest observation per platform and model. Revisions, runtimes and operations can differ; this is not a ranking. Earlier observations remain in the full history.

All questions / TayHCode/tailscale-agent-access · Full history →

Works

README claims checked out on 6 recorded setups. Untested combinations remain unknown.

Answers apply to each recorded setup, revision and operation. Earlier observations remain in Run details.

Works with my model and platform?
PlatformHaikuLuna
Linuxyesyes
Windowsyesyes
macOSyesyes

Linux / Haiku 5.5: yes · Oct 22, 2026 · e43cfd6e93bb · claude-code / 2.1.300 (Claude Code)
Recorded operation: python3 agent_access.py --policy.coldpass/scoped-policy.json generate --output.coldpass/probe-scoped.hujson-checker-e5494a2a69364b7a9f246bcb2b9243c4

Linux / Luna: yes · Oct 22, 2026 · e43cfd6e93bb · codex / codex-cli 0.163.0
Recorded operation: python3 agent_access.py summary

Windows / Haiku 5.5: yes · Oct 22, 2026 · e43cfd6e93bb · claude-code / 2.1.300 (Claude Code)
Recorded operation: 'C:\coldpass\python\python.exe' agent_access.py generate --output '.coldpass\out\probe-fragment.hujson-checker-5cdb27927cf34126aefb0553d589918a'

Windows / Luna: yes · Oct 22, 2026 · e43cfd6e93bb · codex / codex-cli 0.163.0
Recorded operation: 'C:\coldpass\python\python.exe' agent_access.py generate --output 'C:\coldpass\workspace\generated-fragment.hujson-checker-236771112e4644209a6afafad3af22fe'

macOS / Haiku 5.5: yes · Oct 22, 2026 · e43cfd6e93bb · claude-code / 2.1.300 (Claude Code)
Recorded operation: /opt/homebrew/bin/python3 agent_access.py --policy.coldpass/scoped/scoped-policy.json generate

macOS / Luna: yes · Oct 22, 2026 · e43cfd6e93bb · codex / codex-cli 0.163.0
Recorded operation: /opt/homebrew/bin/python3 agent_access.py summary

Does what it says
3 of 3 claims checked out
  • scoped mode grants only the listed LAN host and ports: Real input produced the expected result: <internal-ip>

    Source: Choose `everything` to grant an agent all IP traffic to all destinations, or `scoped` to grant only its listed LAN hosts and ports.

    Input: scoped policy with printer <internal-ip> allowed on tcp:631 for agent-a

    Expected: generated grant lists only <internal-ip> with tcp:631 and udp:8000-8010

    Command: /opt/homebrew/bin/python3 agent_access.py --policy .coldpass/scoped/scoped-policy.json generate

    Observed: <internal-ip>" ], "ip": [ "tcp:631", "udp:8000-8010" ] }, { "src": [ "tag:agent-agent-b" ], "dst": [ "*" ], "ip": [ "*" ] } ], "tests": [ { "src": "tag:agent-agent-a", "proto": "tcp", "accept": [ "<internal-ip>:631" ], "deny": [ "198.51.100.10:631", "tag:agent-agent-b:631" ] }, { "src": "tag:agent-agent-a", "proto": "udp", "accept": [ "<internal-ip>:8000-8010" ], "deny": [ "1

  • a documentation placeholder address is rejected by validation: Real input produced the expected result: must be a private LAN address within RFC1918 IPv4 ranges

    Source: The address above is a documentation placeholder and is intentionally rejected by validation.

    Input: scoped policy using documentation address 192.0.2.25

    Expected: validation exits 2 with a private-address policy error

    Command: /opt/homebrew/bin/python3 agent_access.py --policy .coldpass/scoped/placeholder-policy.json validate

    Observed: must be a private LAN address within RFC1918 IPv4 ranges'

  • diff reviews grant changes without writing: Real input produced the expected result: + {"dst":["<internal-ip>"]

    Source: To review grant changes without writing, run `python3 agent_access.py diff --against existing-policy.json`;

    Input: existing policy with one agent-b grant

    Expected: diff prints the agent-a printer grant as an addition and no removals

    Command: /opt/homebrew/bin/python3 agent_access.py --policy .coldpass/scoped/scoped-policy.json diff --against .coldpass/scoped/existing-policy.json

    Observed: + {"dst":["<internal-ip>"],"ip":["tcp:631","udp:8000-8010"],"src":["tag:agent-agent-a"]} Removed grants: (none)

MCP
No
Local or hosted
local
Local: runs on your machine
Needs keys or browser
None seen
Maintained
Maintained
actively maintained: last commit 7 days ago; 1 commits and 0 merged PRs in 90 days; 0 open issues, 0 open PRs; no releases

Last commit Oct 14, 2026

Setup time
2 min 43 s
2 min 43 s to a working result
Issue
none found in recorded check
Examples
4 of 4 ran as expected
  • quickstart summary lists each agent's grant mode: command completed successfully

    Source: README.md Quickstart

    Input: summary with default policy.json

    Expected: summary header and per-agent grant lines

    Command: /opt/homebrew/bin/python3 agent_access.py summary

    Observed: e not excluded. A range grants the listed TCP/UDP flows to every address it covers; service labels do not restrict destinations. Scoped mode does not model or test ICMP; the everything preset grants all IP traffic, including ICMP. This fragment does not revoke credentials, host permissions, or access provided elsewhere. agent-a (tag:agent-agent-a) Everything preset: grants all IP traffic to all destinations. agent-b (tag:agent-agent-b) Everything preset: grants all IP traffic to all destinations. agent-c (tag:agent-agent-c) Everything preset: grants all IP traffic to all destinations.

  • validate the default policy: command completed successfully

    Source: README.md Quickstart

    Input: default policy.json

    Expected: Policy is valid.

    Command: /opt/homebrew/bin/python3 agent_access.py validate

    Observed: Policy is valid.

  • generate writes the Tailscale policy fragment to a file: command completed successfully

    Source: README.md Quickstart

    Input: generate --output .coldpass/examples/generated-fragment.hujson

    Expected: fragment with tagOwners, grants, tests

    Command: /opt/homebrew/bin/python3 agent_access.py generate --output .coldpass/examples/generated-fragment.hujson-checker-9a98079f2c4e4ddda13d5c9798245474

    Observed: ent-a": [ "autogroup:admin" ], "tag:agent-agent-b": [ "autogroup:admin" ], "tag:agent-agent-c": [ "autogroup:admin" ] }, "grants": [ { "src": [ "tag:agent-agent-a" ], "dst": [ "*" ], "ip": [ "*" ] }, { "src": [ "tag:agent-agent-b" ], "dst": [ "*" ], "ip": [ "*" ] }, { "src": [ "tag:agent-agent-c" ], "dst": [ "*" ], "ip": [ "*" ] } ], "tests": [] }

  • unit test suite (documented in release-check.json): command completed successfully

    Source: release-check.json

    Input: python3 -m unittest discover -s tests

    Expected: 35 tests pass

    Command: /opt/homebrew/bin/python3 -m unittest discover -s tests

    Observed: ................................... ---------------------------------------------------------------------- Ran 35 tests in 18.156s OK

Evidence
README behavior checked
Agent interfaces
CLI
CLI

Observed interfaces were exercised by the checker. Documentation alone does not confirm compatibility.

Tested with
Claude Code / Haiku 5.5
  • macOS: Claude Code / Haiku 5.5 · agent 2.1.300 (Claude Code) · Checked operation passed · Oct 22, 2026 on e43cfd6e93bb

Each entry describes its recorded operation and source revision. Other agent clients were not tested.

Safety
No sign of malicious behavior found in these checks

No problems traced to this repository code in these checks; a short run cannot prove safety.

Full event, path and process lists for each observed setup:

Haiku 5.5 (claude-haiku-5-5)

Checked operation passed.

Operation: python3 agent_access.py --policy .coldpass/scoped-policy.json generate --output .coldpass/probe-scoped.hujson-checker-e5494a2a69364b7a9f246bcb2b9243c4

Recorded scope: product operation

Agent-reported scope: Generate and review Tailscale grant fragments for agent-a/b/c with the offline agent_access.py CLI (summary, validate, generate, diff, and a scoped policy via --policy)

run
TayHCode--tailscale-agent-access-claude-haiku-5-5-1792644847350310105
date
Oct 22, 2026
source commit
e43cfd6e93bb2e58811dcea3f9266dec8b52658e
test version
readme-1924d8f529c6fc4500abb29916f5c9ac942027957e7b82d725d5610040cd3b4a
purpose
Initial run
platform
linux
reasoning / service tier
low / not recorded
recorded resources
4096 MiB RAM, 2 CPUs, 32768 MiB disk
model / version
claude-haiku-5-5 / claude-haiku-5-5
agent runtime / version
claude-code / 2.1.300 (Claude Code)
checker
PASS · cli
Recorded source
https://github.com/TayHCode/tailscale-agent-access
Backend
hyperv
Tool transport
mcp-assigned-vm
Requested source pin (not an observed revision)
{"commit": "e43cfd6e93bb2e58811dcea3f9266dec8b52658e", "url": "https://github.com/TayHCode/tailscale-agent-access"}
time to checked operation
308.3s
total attempt
332.2s
reported cost (usage complete)
$0.016070

Checked operation passed.

Operation: 'C:\coldpass\python\python.exe' agent_access.py generate --output '.coldpass\out\probe-fragment.hujson-checker-5cdb27927cf34126aefb0553d589918a'

Recorded scope: product operation

Agent-reported scope: Generate a Tailscale policy fragment (tagOwners, grants, tests) for agent-a, agent-b, agent-c, plus validate, summary, and diff

run
TayHCode--tailscale-agent-access-claude-haiku-5-5-1792648582398688048
date
Oct 22, 2026
source commit
e43cfd6e93bb2e58811dcea3f9266dec8b52658e
test version
readme-1924d8f529c6fc4500abb29916f5c9ac942027957e7b82d725d5610040cd3b4a
purpose
Initial run
platform
windows
reasoning / service tier
low / not recorded
recorded resources
4096 MiB RAM, 2 CPUs, 32768 MiB disk
model / version
claude-haiku-5-5 / claude-haiku-5-5
agent runtime / version
claude-code / 2.1.300 (Claude Code)
checker
PASS · cli
Recorded source
https://github.com/TayHCode/tailscale-agent-access
Backend
hyperv
Tool transport
mcp-assigned-vm
Requested source pin (not an observed revision)
{"commit": "e43cfd6e93bb2e58811dcea3f9266dec8b52658e", "url": "https://github.com/TayHCode/tailscale-agent-access"}
time to checked operation
169.2s
total attempt
460.1s
reported cost (usage complete)
$0.014097

Checked operation passed.

Operation: /opt/homebrew/bin/python3 agent_access.py --policy .coldpass/scoped/scoped-policy.json generate

Recorded scope: product operation

Agent-reported scope: Generate and review Tailscale grant fragments for three agents with the offline agent_access.py CLI (summary, validate, generate, diff), including scoped mode

run
TayHCode--tailscale-agent-access-claude-haiku-5-5-1792690919570078318
date
Oct 22, 2026
source commit
e43cfd6e93bb2e58811dcea3f9266dec8b52658e
test version
readme-71af40079ee68fdda131463405700e8219a99ea96bf0185ac605d5ecffaf5971
purpose
Test repair rerun
platform
macos
reasoning / service tier
low / not recorded
recorded resources
4096 MiB RAM, None CPUs, 32768 MiB disk
model / version
claude-haiku-5-5 / claude-haiku-5-5
agent runtime / version
claude-code / 2.1.300 (Claude Code)
checker
PASS · cli
Recorded source
https://github.com/TayHCode/tailscale-agent-access
Backend
tart
Tool transport
mcp-assigned-vm
Requested source pin (not an observed revision)
{"commit": "e43cfd6e93bb2e58811dcea3f9266dec8b52658e", "url": "https://github.com/TayHCode/tailscale-agent-access"}
time to checked operation
163.1s
total attempt
198.6s
reported cost (usage complete)
$0.014788

Luna (gpt-6-luna)

Checked operation passed.

Operation: python3 agent_access.py summary

Recorded scope: product operation

Agent-reported scope: Summarize, validate, and generate an offline Tailscale grant policy fragment

run
TayHCode--tailscale-agent-access-gpt-6-luna-1792683707660888275
date
Oct 22, 2026
source commit
e43cfd6e93bb2e58811dcea3f9266dec8b52658e
test version
readme-1002bc876f0002f622e14f9c5b99f9f66721fccc0356f4d53ed24f2f79a3ac5d
purpose
Initial run
platform
linux
reasoning / service tier
low / standard
recorded resources
4096 MiB RAM, 2 CPUs, 32768 MiB disk
model / version
gpt-6-luna / gpt-6-luna
agent runtime / version
codex / codex-cli 0.163.0
checker
PASS · cli
Recorded source
https://github.com/TayHCode/tailscale-agent-access
Backend
hyperv
Tool transport
mcp-assigned-vm
Requested source pin (not an observed revision)
{"commit": "e43cfd6e93bb2e58811dcea3f9266dec8b52658e", "url": "https://github.com/TayHCode/tailscale-agent-access"}
time to checked operation
17.7s
total attempt
34.7s
reported cost
Not reported; not assumed to be zero.

Checked operation passed.

Operation: 'C:\coldpass\python\python.exe' agent_access.py generate --output 'C:\coldpass\workspace\generated-fragment.hujson-checker-236771112e4644209a6afafad3af22fe'

Recorded scope: product operation

Agent-reported scope: Generate an offline Tailscale policy fragment from the checked-in policy.

run
TayHCode--tailscale-agent-access-gpt-6-luna-1792684932920431201
date
Oct 22, 2026
source commit
e43cfd6e93bb2e58811dcea3f9266dec8b52658e
test version
readme-1002bc876f0002f622e14f9c5b99f9f66721fccc0356f4d53ed24f2f79a3ac5d
purpose
Initial run
platform
windows
reasoning / service tier
low / standard
recorded resources
4096 MiB RAM, 2 CPUs, 32768 MiB disk
model / version
gpt-6-luna / gpt-6-luna
agent runtime / version
codex / codex-cli 0.163.0
checker
PASS · cli
Recorded source
https://github.com/TayHCode/tailscale-agent-access
Backend
hyperv
Tool transport
mcp-assigned-vm
Requested source pin (not an observed revision)
{"commit": "e43cfd6e93bb2e58811dcea3f9266dec8b52658e", "url": "https://github.com/TayHCode/tailscale-agent-access"}
time to checked operation
47.7s
total attempt
127.4s
reported cost
Not reported; not assumed to be zero.

Checked operation passed.

Operation: /opt/homebrew/bin/python3 agent_access.py summary

Recorded scope: product operation

Agent-reported scope: Summarize configured Tailscale grants

run
TayHCode--tailscale-agent-access-gpt-6-luna-1792689884767893169
date
Oct 22, 2026
source commit
e43cfd6e93bb2e58811dcea3f9266dec8b52658e
test version
readme-0f3d0bf16b16cf3f0e7a538608f68c653dedb635064823cab4ef3c773fbbd53f
purpose
Test repair rerun
platform
macos
reasoning / service tier
low / standard
recorded resources
4096 MiB RAM, None CPUs, 32768 MiB disk
model / version
gpt-6-luna / gpt-6-luna
agent runtime / version
codex / codex-cli 0.163.0
checker
PASS · cli
Recorded source
https://github.com/TayHCode/tailscale-agent-access
Backend
tart
Tool transport
mcp-assigned-vm
Requested source pin (not an observed revision)
{"commit": "e43cfd6e93bb2e58811dcea3f9266dec8b52658e", "url": "https://github.com/TayHCode/tailscale-agent-access"}
time to checked operation
27.5s
total attempt
67.1s
reported cost
Not reported; not assumed to be zero.

More Coldpass questions