The control plane for
auto-research work.
arcp orchestrates a fleet of agents — code improvers, paper-to-code pipelines, thesis explorers — over a single API and a single UI. Submit a use case, watch the events stream, review the result.
What you can do
Each card is a research workflow. Behind each one is a job kind, a spec schema, and one or more sibling agents that arcp orchestrates.
Improve a codebase
Run autocode-sdk against a repo to refactor, add tests, or apply a research recipe. Local by default; toggle on cloud sandboxing when isolation matters.
Sweep all bugs
Discover and fix bugs across the tree in parallel, one job per finding. Aggregates into a single review-ready report.
Comprehend a repo
Build a repomap, code graph, and embeddings before you touch anything. Cheap, deterministic, no LLMs required.
Paper → code
Hand the paperToCode agent a PDF or arXiv ID. It paces itself through extraction, planning, drafting, and verification — pausing for your approval at the gates that matter.
Code → paper
Take a thesis, experiments, and benchmarks and produce the appropriate publishable output — blog, whitepaper, research paper, or executable notebook.
Explore a thesis
Search-driven literature scan: pulls related work, ranks claims, and produces a structured map you can navigate before committing to a direction.
Benchmark inference
Send a model + dataset through InferBench and collect latency, throughput, and quality metrics. A/B two variants with a judge model when you need a head-to-head.
Fine-tune a model
Point autotune at a base model and dataset and pick a method — LoRA, QLoRA, SFT, or DPO. Get back a trained adapter plus training metrics, GPU work routed through Tensile.
How it's wired
arcp
API + jobs + UI. Holds the queue, the journal, and the orchestration logic. The piece you're looking at.
autocode-sdk
The code agent. Runs locally by default for low-friction loops. Sandbox mode (Modal / Docker / E2B) is opt-in.
paperToCode
Paper → code pipeline as its own service. arcp talks to it over HTTP, same shape as the scraping agent.
Tensile — the Research Commons infrastructure layer.
Schedulers, sandboxes, observability, and shared services that every RC product runs on. arcp is one of those products.