---
title: "Give Your Coding Agents a Memory You Own"
description: "funes is presented as a durable memory layer for coding agents that turns existing session traces into searchable, provenance-backed working memory across machines and agent types. A single binary adds recall and get tools to Claude Code, Codex, pi, or Hermes, builds an incremental local index, and automatically indexes completed turns. Its deterministic pipeline normalizes traces, chunks them, embeds them with a pinned local model, stores them in Lance, and combines vector and BM25 search with rank fusion, cross-encoder reranking, recency weighting, and neighboring chunks. Memories can remain local or be bound to a private Hugging Face dataset owned by the user, with credential redaction, secret scanning, local caching, and Hub-provided access control and versioning. In the handoff-vs-recall benchmark, recall was the cheapest option on both tasks, costing 8x less than a written handoff on one and 4x less on the other, while compaction succeeded on only one task."
---

# Give Your Coding Agents a Memory You Own

[Hugging Face](https://yomu.fyi/company/hugging-face) · David Corvoysier · Sep 3, 2026

**Type:** Announcement

## Summary

funes is presented as a durable memory layer for coding agents that turns existing session traces into searchable, provenance-backed working memory across machines and agent types. A single binary adds recall and get tools to Claude Code, Codex, pi, or Hermes, builds an incremental local index, and automatically indexes completed turns. Its deterministic pipeline normalizes traces, chunks them, embeds them with a pinned local model, stores them in Lance, and combines vector and BM25 search with rank fusion, cross-encoder reranking, recency weighting, and neighboring chunks. Memories can remain local or be bound to a private Hugging Face dataset owned by the user, with credential redaction, secret scanning, local caching, and Hub-provided access control and versioning. In the handoff-vs-recall benchmark, recall was the cheapest option on both tasks, costing 8x less than a written handoff on one and 4x less on the other, while compaction succeeded on only one task.

## Context

Coding agents lose the reasoning, findings, errors, and decisions from completed sessions, especially when work moves between machines or agents. Session logs preserve this history, but they remain an archive without indexing, retrieval, ranking, and exact provenance.

## Approach / What changed

funes is a single binary that parses supported traces into a common turn-and-block structure, incrementally indexes them locally, and provides recall and get tools through agent integrations. It combines vector and BM25 retrieval with fusion, cross-encoder reranking, recency weighting, and neighboring chunks. A memory can also be published and synchronized through a private Hugging Face dataset, with credential redaction, secret scanning, and local caching.

## Takeaways

- funes stores raw evidence rather than distilling facts at write time; recall identifies the agent, timestamp, session, and turn, while get opens the full turn and surrounding context.
- funes add installs an agent integration and incremental indexing, whereas funes ask is read-only, answers one question, and does not change the agent’s persistent setup.
- Remote memories use private-by-default Hugging Face datasets; credentials are redacted during indexing, and publishing scans each chunk again for content that still looks like a secret.

**Tags:** [AI Agents](https://yomu.fyi/topic/ai-agents), [Open Source](https://yomu.fyi/topic/open-source), [Search](https://yomu.fyi/topic/search)

- Source: [Hugging Face](https://huggingface.co/blog/funes)
- Source URL: https://huggingface.co/blog/funes
- Ingested by Yomu: 2026-09-03T11:17:45.595Z

[Read original post](https://huggingface.co/blog/funes)