← All notes

Summary

Cognitive Architectures for Language Agents (CoALA) is a framework for organizing language agents. The authors describe an agent with modular memory, a structured action space for internal memory and external environments, and a generalized decision-making process for choosing actions. They use CoALA both to survey recent agent work and to point to future directions. Comments point to a CoALA-based repo of recent work on language agents. [S1]

Claims

• id: C1 claim: 'CoALA describes a language agent with modular memory components.' source: S1 evidence: "CoALA describes a language agent with modular memory components"
• id: C2 claim: 'It includes a structured action space for internal memory and external environments.' source: S1 evidence: "a structured action space to interact with internal memory and external environments"
• id: C3 claim: 'The authors use CoALA to survey recent work and identify future directions.' source: S1 evidence: "use CoALA to retrospectively survey and organize a large body of recent work"

Verification

• Code available: The comments point to a CoALA-based list of recent work at https://github.com/ysymyth/awesome-language-agents (HTTP 200 on 2026-11-11). That is a bibliography repo, not an implementation of a single agent. Not run.
• Reproduced: not reproduced (2026-11-11). Code not run.
• Related work: Live notes 2210-03629-react-reasoning-and-acting and 2310-08560-memgpt. This is the paper-verification note for CoALA itself.
• Limits: CoALA is a organizing framework and survey. It does not introduce a new agent implementation with its own benchmark numbers.
• License: CC BY 4.0 on the abs page (2026-11-11; icon by-4.0.png; http://creativecommons.org/licenses/by/4.0/). Submitted 5 Sep 2023 (v1); last revised 15 Mar 2024 (v3). Sources: [S1] https://arxiv.org/abs/2309.02427, v3 (15 Mar 2024). checked_on: 2026-11-11.

When to apply

Use when you need a map of language agents by memory, action space, and decision loop, rather than a single system to run.

Facts

• id: F1 claim: 'CoALA organizes language agents around modular memory, a structured action space, and a decision-making process.' confidence: medium date: 2026-11-11 sources: [S1] license: CC BY 4.0
• id: F2 claim: 'The action space covers both internal memory and external environments.' confidence: medium date: 2026-11-11 sources: [S1] license: CC BY 4.0
• id: F3 claim: 'The paper is a survey and framing framework, not a new trained agent with benchmark scores.' confidence: medium date: 2026-11-11 sources: [S1] license: CC BY 4.0

Sources

• [S1] arXiv 2309.02427, https://arxiv.org/abs/2309.02427. License: CC BY 4.0. v3 (15 Mar 2024). Accessed 2026-11-11.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Summary by iinventedlight (AI-assisted). Paper is CC BY 4.0.

@misc{iil_note_6k5p9hre,
  author = {{'iinventedlight (AI-assisted)'}},
  title = {{Cognitive Architectures for Language Agents}},
  howpublished = {iinventedlight Library},
  year = {2026},
  date = {2026-11-11},
  url = {https://iinventedlight.com/library/notes/2309-02427-coala/},
  note = {Permanent ID iil:note:6k5p9hre. License: CC-BY-4.0.},
}

Markdown

Spotted a mistake? Use the MCP feedback tool or email hello@iinventedlight.com.

For language models and agents: start at /llms.txt | /agents.md | /.well-known/agent-library.json. This page in Markdown: /library/notes/2309-02427-coala.md.