agent–computer interface
In plain wordsThe commands, tools and feedback designed for an AI agent to work with a computer — the agent's version of a user interface, shaped around how language models read and act.
The set of actions, tool definitions, feedback formats and documentation through which an AI agentAI agent, n.: An AI system that does not just answer but acts: given a goal, it decides on steps, uses tools such as a browser or code editor, checks the results and keeps going until the job is done. operates a computer environment, treated as a design object in its own right — analogous to the human–computer interface but built for language models. Good ACIs offer a small number of simple, well-documented commands; return concise, informative feedback (for instance, showing the lines around an edit and flagging syntax errors at once); guard against common mistakes (requiring absolute file paths, rejecting edits that break syntax); and keep observations short enough not to flood the context windowcontext window, n.: How much an AI model can read and keep in view at once: the instructions, the conversation so far and any attached documents. Text that does not fit has to be cut, summarized or left out.. The concept, introduced with SWE-agentSWE-agent, prop. n.: An open-source research system from Princeton that turns a language model into a software-fixing agent by giving it a specially designed set of commands for browsing, editing and testing code., showed that interface design alone can change an agent's success rate substantially with the same model, and it underlies current advice to invest as much effort in tool descriptions as in prompts.
Swapping our raw shell for an agent–computer interface with a search command and a linting editor fixed most failed runs.
Origin
Introduced by John Yang, Carlos E. Jimenez, Ofir Press and colleagues at Princeton in the SWE-agent paper (May 2024; NeurIPS 2024), which argued that language-model agents are a new category of end user deserving purpose-built interfaces. Anthropic's 'Building effective agents' (December 2024) spread the idea, recommending 'poka-yoke' tool designs that make mistakes harder.
Usage note
Narrower than harnessharness, n.: The software around an AI model that gives it instructions and tools, carries out its actions and decides when it stops. The model is the engine; the harness is the rest of the car. design: the ACI is what the agent sees and can do, not the loop that runs it.
Related
- SWE-agentSWE-agent, prop. n.: An open-source research system from Princeton that turns a language model into a software-fixing agent by giving it a specially designed set of commands for browsing, editing and testing code.
- tooltool, n.: Any outside capability an AI model is allowed to use — a web search, a calculator, a code runner, a database or an app — described to the model so it can ask for it by name.
- harnessharness, n.: The software around an AI model that gives it instructions and tools, carries out its actions and decides when it stops. The model is the engine; the harness is the rest of the car.
- coding agentcoding agent, n.: An AI agent that writes and fixes software by itself: it reads a codebase, edits files, runs commands and tests, and keeps working until a feature is built or a bug is fixed.
- computer usecomputer use, n.: An AI's ability to work a computer through its normal screen and controls: it studies pictures of what is displayed and sends mouse and keyboard actions, so it can use ordinary apps that were built for people.
- SWE-benchSWE-bench, prop. n.: A test that asks AI to fix real bugs and add features in real open-source Python projects, checked by running each project's own tests; the best-known yardstick for AI coding agents.
Sources
- Yang, J., Jimenez, C. E., Wettig, A., Lieret, K., Yao, S., Narasimhan, K. & Press, O. (2024). SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering. NeurIPS 2024. arXiv:2405.15793. arxiv.org/abs/2405.15793
- Schluntz, E. & Zhang, B. (2024). Building effective agents. Anthropic. anthropic.com/research/building-effective-agents
Mentioned in
computer usecomputer use, n.: An AI's ability to work a computer through its normal screen and controls: it studies pictures of what is displayed and sends mouse and keyboard actions, so it can use ordinary apps that were built for people., code executioncode execution, n.: Letting an AI write a small program and actually run it — to do math, analyze a spreadsheet or test its own code — then use the real output instead of guessing., tooltool, n.: Any outside capability an AI model is allowed to use — a web search, a calculator, a code runner, a database or an app — described to the model so it can ask for it by name., SWE-agentSWE-agent, prop. n.: An open-source research system from Princeton that turns a language model into a software-fixing agent by giving it a specially designed set of commands for browsing, editing and testing code.