The lack of HTTP to reduce Claudecode costs

[KEVEN11-cmd]

I got very excited, even though I spent $28 on the API and lost $20 on the same day. The project itself is quite encouraging, and I'm really excited for updates so I can use it on a VPS and, of course, use HTTP to reduce costs and keep the company running 24/7. I hope this happens as soon as possible.

[ttomiczek]

Can you try to make sense here? How would a lack of HTTP change anything? Paperclip is a work manager essentially - not an LLM. HTTP does not really enter anywhere.

[jebilbrey]

I think he's referring to the HTTP that shows up under Adapter type so that you can point an agent to something like Ollama or LM Studio locally.

[ttomiczek]

Which still makes no sense because ht talks about costs. For what? The "only" cost you normally have is API - so this literally makes no sense.

[jebilbrey]

Not if you use local LLM, which is the point of using Ollama or LM Studio running a local model. Jason

…

On Sat, Apr 11, 2026 at 6:13 AM ttomiczek - ***@***.*** ***@***.***> wrote: Which still makes no sense because ht talks about costs. For what? The "only" cost you normally have is API - so this literally makes no sense. — Reply to this email directly, view it on GitHub <#3229 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACXL7PQUJ2CDR72SKJX7R3D4VISE5AVCNFSM6AAAAACXSYY5RSVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTMNJSGY2DMMQ> . You are receiving this because you commented.Message ID: ***@***.***>

[ttomiczek]

Again, makes no sense because then he would not complain about api costs at all.

…

On Sat, Apr 11, 2026 at 4:42 PM jebilbrey ***@***.***> wrote: Not if you use local LLM, which is the point of using Ollama or LM Studio running a local model. Jason On Sat, Apr 11, 2026 at 6:13 AM ttomiczek - ***@***.*** ***@***.***> wrote: > Which still makes no sense because ht talks about costs. For what? The > "only" cost you normally have is API - so this literally makes no sense. > > — > Reply to this email directly, view it on GitHub > < #3229 (comment)>, > or unsubscribe > < https://github.com/notifications/unsubscribe-auth/ACXL7PQUJ2CDR72SKJX7R3D4VISE5AVCNFSM6AAAAACXSYY5RSVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTMNJSGY2DMMQ> > . > You are receiving this because you commented.Message ID: > ***@***.***> > — Reply to this email directly, view it on GitHub <#3229 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/B7LZGDY46ZBCGT4MMA56UEL4VI4T5AVCNFSM6AAAAACXSYY5RSVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTMNJSGY4DMNY> . You are receiving this because you commented.Message ID: ***@***.***>

[jebilbrey]

If you read what he wrote, he mentioned API costs but then said he’s excited for HTTP when it becomes available. Which I interpret as him wanting to switch off cloud API costs and going to a local LLM agent. Jason On Sat, Apr 11, 2026 at 7:46 AM ttomiczek - ***@***.*** ***@***.***> wrote:

…

Again, makes no sense because then he would not complain about api costs at all. On Sat, Apr 11, 2026 at 4:42 PM jebilbrey ***@***.***> wrote: > Not if you use local LLM, which is the point of using Ollama or LM Studio > running a local model. > > Jason > > > On Sat, Apr 11, 2026 at 6:13 AM ttomiczek - ***@***.*** > ***@***.***> wrote: > > > Which still makes no sense because ht talks about costs. For what? The > > "only" cost you normally have is API - so this literally makes no sense. > > > > — > > Reply to this email directly, view it on GitHub > > < > #3229 (comment)>, > > > or unsubscribe > > < > https://github.com/notifications/unsubscribe-auth/ACXL7PQUJ2CDR72SKJX7R3D4VISE5AVCNFSM6AAAAACXSYY5RSVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTMNJSGY2DMMQ> > > > . > > You are receiving this because you commented.Message ID: > > ***@***.***> > > > > — > Reply to this email directly, view it on GitHub > < #3229 (comment)>, > or unsubscribe > < https://github.com/notifications/unsubscribe-auth/B7LZGDY46ZBCGT4MMA56UEL4VI4T5AVCNFSM6AAAAACXSYY5RSVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTMNJSGY4DMNY> > . > You are receiving this because you commented.Message ID: > ***@***.***> > — Reply to this email directly, view it on GitHub <#3229 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACXL7PT7CYJPTEPSQBCMXEL4VI5A3AVCNFSM6AAAAACXSYY5RSVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTMNJSGY4DQMY> . You are receiving this because you commented.Message ID: ***@***.***>

[robcoward]

How about you configure your 'agents' to use the OpenCode adapter? Then you can configure OpenCode to connect to whatever provider you want. Personally, I have mine connecting to a LiteLLM instance, which I then use to route requests wherever I want, ollama running on a pc with eGPU, LM Studio on a M4 Max Macbook, or to other cheaper cloud providers like Moonshot, Minimax, Groq etc

[nkaramesinis]

Wow that's a smart approach! Thanks for the heads up!