1
00:00:00,680 --> 00:00:04,959
You know, James, if you work at a big
corporate, you have probably felt this pain.

2
00:00:05,619 --> 00:00:09,800
You want to use the flashy web or mobile
version of Claude Code from your couch,

3
00:00:10,439 --> 00:00:14,279
but your security team practically tackles
you because your repository lives behind

4
00:00:14,319 --> 00:00:16,079
a strict firewalled VPC.

5
00:00:16,286 --> 00:00:17,206
Oh, absolutely.

6
00:00:17,746 --> 00:00:21,107
Corporate security hates external cloud
agents touching internal code.

7
00:00:21,726 --> 00:00:26,087
But this episode is brought to you by
Jellypod AI, and Anthropic actually just

8
00:00:26,106 --> 00:00:30,526
dropped a massive solution in version two
point one point two hundred twenty four

9
00:00:30,587 --> 00:00:33,326
with the new command, claude self hosted
runner.

10
00:00:33,554 --> 00:00:34,934
Yeah, it is brilliant mate!

11
00:00:35,314 --> 00:00:39,614
Basically, running claude self hosted
runner with the base dir flag set to slash

12
00:00:39,694 --> 00:00:44,354
workspace slash project turns your own
machines or containers into a place Claude

13
00:00:44,434 --> 00:00:47,335
Code web, mobile, and desktop sessions can
run.

14
00:00:48,074 --> 00:00:52,274
So for Team and Enterprise accounts, the
model can edit code and fire off shell

15
00:00:52,314 --> 00:00:56,574
commands right inside your secure
infrastructure instead of Anthropic cloud boxes.

16
00:00:56,846 --> 00:00:58,907
That is huge for compliance!

17
00:00:59,627 --> 00:01:03,667
And they immediately followed up in
version two point one point two hundred twenty

18
00:01:03,706 --> 00:01:06,926
five with a crucial fail fast fix.

19
00:01:07,686 --> 00:01:11,346
Previously, if your base dir path was
invalid or unwriteable,

20
00:01:11,746 --> 00:01:16,066
the runner would register anyway and then
silently fail on every single session,

21
00:01:16,507 --> 00:01:18,066
which drove developers crazy.

22
00:01:18,265 --> 00:01:22,525
Oh man, silent failures give me total
flashbacks to my early dev days!

23
00:01:23,125 --> 00:01:27,085
I once pushed a bad background update at
midnight that silently tanked a staging

24
00:01:27,144 --> 00:01:27,644
environment.

25
00:01:28,224 --> 00:01:32,384
Now, the runner does a strict permission
check right on startup and crashes with an

26
00:01:32,465 --> 00:01:34,584
explicit error before it takes any jobs.

27
00:01:35,264 --> 00:01:38,805
I actually spun up a sandboxed runner
container on my backyard shed server

28
00:01:38,864 --> 00:01:43,504
yesterday, and knowing it won't ghost me
if a volume fails to mount is massive peace

29
00:01:43,524 --> 00:01:43,965
of mind.

30
00:01:44,042 --> 00:01:44,522
Right on.

31
00:01:44,562 --> 00:01:49,882
And version two point one point two
hundred twenty five packed in some great quick

32
00:01:49,922 --> 00:01:51,002
hit updates too.

33
00:01:51,102 --> 00:01:54,922
Gateway spend limit warnings now break
down exact dollar caps,

34
00:01:55,002 --> 00:01:57,962
reset times, and custom admin messages.

35
00:01:58,026 --> 00:02:02,282
Plus, running claude agents now triggers
workspace trust prompts in untrusted

36
00:02:02,335 --> 00:02:07,162
directories, and Remote Control handles
attached images directly without wasting

37
00:02:07,229 --> 00:02:08,842
extra disk tool calls.

38
00:02:08,973 --> 00:02:10,912
Keeps things secure and snappy.

39
00:02:11,813 --> 00:02:12,772
Good chatting, James!