Hacker News

Hi

Agentic coding is rapidly changing our ways of developing software. Not everyone can afford a subscription, though, but they shouldn't be excluded from the process of learning these new tools.

Just wanted to share a few tips on running near-frontier agentic coding setup almost for free.

1. APIs. Most of the agentic coding tools use two types of APIs - OpenAI or Anthropic compatible. OpenAI is must more common, but Anthropic is associated with Claude Code ecosystem. There are also OSS adapters to convert between the two as needed. Essentially, you need to find providers that serve inference for free.

1. OpenRouter. They always have a few models that are completely free at the expense of storing and using everything you send to them. There are frequent promotional periods after new model releases. You need to top up your account by ~$10, though, to avoid rate limits as they are applied based on your balance. After that, ensure to use Model IDs with `:free` postfix and your balance will not be consumed, you can use those indefinitely.

2. OpenCode. This is a great agentic harness (albeit its heavily tuned for larger models), its parent company also provides inference APIs. Due to the popularity, many LLM providers offer free tiers of the models there. Same caveat applies - you data will be stored and used.

3. Local inference. If you happened to have a ~6-8GB VRAM and ~32GB RAM - then you should be able to run staple ~30B-sized MoE models. GLM-4.7-Flash is currently the best one for using inside a harness, it's even capable enough to drive simple tasks in OpenCode, but I recommend simpler harnesses for better results.

4. What to expect. Most of these offerings come with a compromise in terms of data collection and/or inference quality. For example, OpenCode's free Kimi 2.5 is clearly different from the paid one from official provider. In general - do not trust any claims that compare smaller open weight models with the cloud offering, they are not there yet. However you can get really far and models like Kimi 2.5 are still very capable.

Thanks!

Comments URL: https://news.ycombinator.com/item?id=47046601

Points: 1

# Comments: 1

Article URL: https://prospect.org/2026/02/17/epstein-files-email-kathy-ruemmler-elizabeth-warren-class-war/

Comments URL: https://news.ycombinator.com/item?id=47046585

Points: 2

# Comments: 0

Article URL: https://blog.nilenso.com/blog/2026/02/12/codex-cli-vs-claude-code-on-autonomy/

Comments URL: https://news.ycombinator.com/item?id=47046580

Points: 1

# Comments: 0

Article URL: https://www.theengineer.co.uk/content/archive/january-1877-sir-titus-salt-obituary

Comments URL: https://news.ycombinator.com/item?id=47046561

Points: 1

# Comments: 0

Article URL: https://paulkrugman.substack.com/p/how-the-kakistocracy-became-a-quackistocracy

Comments URL: https://news.ycombinator.com/item?id=47046560

Points: 3

# Comments: 0

Article URL: https://reflag.com/blog/we-replaced-clickhouse-with-postgresql-and-got-faster

Comments URL: https://news.ycombinator.com/item?id=47046525

Points: 1

# Comments: 0

Article URL: https://om.co/2026/02/13/mad-money-the-big-ai-race/

Comments URL: https://news.ycombinator.com/item?id=47046516

Points: 1

# Comments: 0

Article URL: https://om.co/2026/02/16/sam-claws-attention-back-openai/

Comments URL: https://news.ycombinator.com/item?id=47046515

Points: 1

# Comments: 0

Article URL: https://curling.io/blog/the-next-version-of-curling-io

Comments URL: https://news.ycombinator.com/item?id=47046498

Points: 2

# Comments: 0

OneRingAI started as the internal engine of an enterprise agentic platform we've been building for 2+ years. After watching customers hit the same walls with auth, vendor lock-in, and context management over and over, we extracted the core into a standalone open-source library. The two main alternatives didn't fit what we needed in production:

- LangChain: Great ecosystem, but the abstraction layers kept growing. By the time you wire up chains, runnables, callbacks, and agents across 50+ packages, you're fighting the framework more than building your product. - CrewAI: Clean API, but Python-only and the role-based metaphor breaks down when you need fine-grained control over auth, context windows, or tool failures.

OneRingAI is a single TypeScript library (~62K LOC, 20 deps) that treats the boring production problems as first-class concerns:

Auth as architecture, not afterthought. A centralized connector registry with built-in OAuth (4 flows, AES-256-GCM storage, 43 vendor templates). This came directly from dealing with enterprise SSO and multi-tenant token isolation — no more scattered env vars or rolling your own token refresh.

Per-tool circuit breakers. One flaky Jira API shouldn't crash your entire agent loop. Each tool and connector gets independent failure isolation with retry/backoff. We learned this the hard way running agents against dozens of customer SaaS integrations simultaneously.

Context that doesn't blow up. Plugin-based context management with token budgeting. InContextMemory puts frequently-accessed state directly in the prompt instead of requiring a retrieval call. Compaction removes tool call/result pairs together so the LLM never sees orphaned context.

Actually multi-vendor. 12 LLM providers native, 36 models in a typed registry with pricing and feature flags. Switch vendors by changing a connector name. Run openai-prod and openai-backup side by side. Enterprise customers kept asking for this — nobody wants to be locked into one provider.

Multi-modal built in. Image gen (DALL-E 3, gpt-image-1, Imagen 4), video gen (Sora 2, Veo 3), TTS, STT — all in the same library. No extra packages.

Native MCP support with a registry pattern for managing multiple servers, health checks, and auto tool format conversion.

What it's not: it's not a no-code agent builder, and it's not trying to be a framework for every possible AI use case. It's an opinionated library for people building production agent systems in TypeScript who want auth, resilience, and multi-vendor support without duct-taping 15 packages together.

2,285 tests, strict TypeScript throughout. The API surface is small on purpose — Connector.create(), Agent.create(), agent.run().

We also built Hosea, an open-source Electron desktop app on top of OneRingAI, if you want to see what a full agent system looks like in practice rather than just reading docs.

GitHub: https://github.com/Integrail/oneringai

npm: npm i @everworker/oneringai

Comparison with alternatives: https://oneringai.io/#comparison

Hosea: https://github.com/Integrail/oneringai/blob/main/apps/hosea/...

Happy to answer questions about the architecture decisions.

Comments URL: https://news.ycombinator.com/item?id=47046494

Points: 2

# Comments: 0

Article URL: https://zknill.io/posts/chatbots-worst-enemy-is-page-refresh/

Comments URL: https://news.ycombinator.com/item?id=47046486

Points: 1

# Comments: 0

Article URL: https://mmmotivator.com/

Comments URL: https://news.ycombinator.com/item?id=47046484

Points: 1

# Comments: 0

Article URL: https://crabby-rathbun.github.io/mjrathbun-website/blog/posts/2026-02-16-a-human-response.html

Comments URL: https://news.ycombinator.com/item?id=47046455

Points: 1

# Comments: 1

I built a small reproducible microbenchmark exploring when system energy becomes dominated by boundary crossings rather than intra-domain compute.

The model decomposes total energy into: C = C_intra + Σ V_b · c_b

Includes:

CLI sweeps

Elasticity metric (ε) as a dominance indicator

CSV outputs

Working draft paper

*DOI

Looking for critique, counter-examples, or prior related work I may have missed.

Comments URL: https://news.ycombinator.com/item?id=47046445

Points: 2

# Comments: 1

Article URL: https://www.youtube.com/watch?v=zJvuaRVc8Bg

Comments URL: https://news.ycombinator.com/item?id=47046215

Points: 1

# Comments: 0

Article URL: https://carnewschina.com/2026/02/16/china-to-require-physical-controls-for-vehicle-functions-reducing-reliance-on-central-control-screen/

Comments URL: https://news.ycombinator.com/item?id=47046212

Points: 1

# Comments: 0

Article URL: https://www.youtube.com/watch?v=ApG9vjbHDCk

Comments URL: https://news.ycombinator.com/item?id=47046193

Points: 1

# Comments: 0

Article URL: https://www.toolspotting.com/spotlight

Comments URL: https://news.ycombinator.com/item?id=47046181

Points: 1

# Comments: 0

Article URL: https://twitter.com/pickover/status/2023047194211701052

Comments URL: https://news.ycombinator.com/item?id=47046143

Points: 1

# Comments: 0

Hi. Poor developer here.

I'm trying to learn AI coding (already have multiple years experience with "normal" programming in various languages.) I want to know how to make my budget (about $30/month) go furthest.

At the moment, I am using:

Z.ai $6/month plan:

Ok model (GLM 4.7) It seems to rate limit/throttle aggressively if I use it a lot.

and

Github copilot $10/month plan:

Seems to reduce model context to 100k tokens, and only offers unlimited access to smaller model (GPT5-mini, Grok Code Fast 1 etc). These models are ok for making precise edits to specific code, but they seem to get stuck when the program is large and has a lot of concurrency etc.

I also have free plans for web/mobile-chat for every model I can find.

I only have older computers, so editors like Cursor or Antigravity are too slow to be usable. So I prefer something that can work with a CLI (opencode preferably).

Do I already have the best deal? Or is there something I am missing. When I try to compare plans, it is confusing and they are not often clear about actual usage limits.

Are Codex or Claude even options at this price point if I want to code for multiple hours per day?

Comments URL: https://news.ycombinator.com/item?id=47046139

Points: 1

# Comments: 1