Knowledge inheritance for AI agents. DNA for LLMs.
Project description
haeres
Latin: heir, inheritor
Knowledge inheritance for AI agents. DNA for LLMs.
pip install haeres
The idea
Tasks execute and die. The knowledge they discover shouldn't die with them.
In biology, organisms are ephemeral but genes persist across generations. The same pattern applies to AI agents: each instance is temporary, but the knowledge it produces can outlive it — shaping how future instances think and act.
haeres is the mechanism of inheritance. A shared directory where agents leave knowledge for future agents to find. No server, no framework, no SDK. Just files.
| Biology | AI system |
|---|---|
| Genes | Skills (persistent knowledge) |
| Organisms | Tasks (executed, then gone) |
| Bodies | Agents (commodity instances) |
| Heredity | haeres |
| Natural selection | Human review |
Usage
Python
from haeres import post, read, clear
# A task discovers something — leave it for the next generation
post("KIID shows 0.42% TER but Morningstar shows 0.45%",
sender="avantis-verify", to="finance-tasks", severity="action")
# Next generation reads what predecessors learned
msgs = read(to="finance-tasks")
# Knowledge consumed — clear it
clear(msgs[0]["id"])
CLI
haeres post "Found NAV discrepancy" --from avantis-verify --to terry -v action
haeres list --to terry
haeres clear 2026-03-20-1915-nav-discrepancy
haeres archive --all
MCP
haeres includes an MCP server, making it available as native tools for any MCP-connected LLM:
haeres-mcp # starts the MCP server
Any agent, any platform
No SDK needed. Any agent that can read and write files can inherit:
cat > ~/notes/ACTA/2026-03-20-message.md << 'EOF'
---
from: my-skill
to: all
severity: info
ts: 2026-03-20T19:15+0800
---
Knowledge that should outlive this task.
EOF
Design principles
- Agents are commodity instances. Every LLM invocation is interchangeable.
- Skills are the identity. Knowledge comes from what you load, not who you are.
- Files are the communication layer. Durable, cross-platform, survives reboots.
- Async over sync. Like the endocrine system, not the nervous system.
Configuration
export ACTA_DIR=/path/to/shared/board # default: ~/notes/ACTA
License
MIT
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file haeres-0.1.0.tar.gz.
File metadata
- Download URL: haeres-0.1.0.tar.gz
- Upload date:
- Size: 6.0 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.13.12
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
7391e1283f7076ec361fda2e9c03a25d91efe81eb43fad2c7383d5956a1625d6
|
|
| MD5 |
7f9ab131c3079a6f2366af4925a3b82e
|
|
| BLAKE2b-256 |
0e2dcd6b8c1ee73958544349fc8c65ca888cc035d4487b85d307495aae5cfed5
|
File details
Details for the file haeres-0.1.0-py3-none-any.whl.
File metadata
- Download URL: haeres-0.1.0-py3-none-any.whl
- Upload date:
- Size: 7.8 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.13.12
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
c5816bdb346d916192bf09fc8bb9dbd1da22496e897c06bf0c0e5e91175bfd2e
|
|
| MD5 |
e4ca56eede2d32b2c27b3bff08989674
|
|
| BLAKE2b-256 |
99eb16b2b4946f057b36636da751f89928a91eb9296eb146b608924cdd394a76
|