kontext

📄

Declarative

What agents follow is declared in the repository. One Markdown file per decision says what holds, since when and for which paths; .ai/kontext.toml says how it is served. Nothing lives in a service or a UI.

🔁

Reproducible

The same commit gives every teammate and every agent the same team knowledge. The local index is derived and can be deleted; the next call rebuilds it. Check out an older commit and agents see the decisions that held at that commit.

↩️

Reviewed and reversible

Agents capture into a local inbox; a capture becomes team knowledge only through a commit and a pull request. A reversed decision is superseded, not deleted, and git revert undoes a bad change like any other.

🧭

One brief, every agent

ctx_brief gives Claude Code, Codex, Cursor, OpenCode or any MCP client the decisions for the paths it names, within a token budget (1400 by default). Paths in a submodule or another worktree are answered by the repository that owns them.

Several repositories

⏳

Stale knowledge surfaces

An entry is written once while its code keeps moving. One whose paths saw 20 or more commits since someone last looked (configurable) is flagged in the brief, and kontext drift lists what a branch left behind.

Freshness

🔒

Local, no service

One Rust binary is the MCP server, the CLI and the git hooks. No account and no server; nothing leaves the machine unless you configure it.

Declare it

.ai/decisions/2026-09-28-prices-are-integer-cents.md.ai/kontext.toml

markdown

---
id: 2026-09-28-prices-are-integer-cents
kind: decision
title: Prices are integer cents
status: accepted
date: 2026-09-28
paths: [src/billing/**]
author: Jane Doe
---
Floats broke VAT rounding on invoices with many lines.
Every amount is stored and computed as an integer number of cents.

toml

# committed, reviewed like code
[brief]
budget_tokens = 1400     # what one brief may cost an agent
max_decisions = 12

[freshness]
threshold_commits = 20   # flag an entry after this many commits on its paths

[hooks]
validate = true          # malformed entries and secrets block the commit
trailers = true          # `Decision: <id>` on the commit that ships a decision

[secrets]
scan = "staged"          # scan every staged file, not only knowledge

From these two files, kontext derives the rest:

an agent that calls ctx_brief with focus: ["src/billing"] gets the decision first,

kontext why src/billing/round.ts:40 shows it next to the blame and the commits behind it,

the commit that ships it carries Decision: 2026-09-28-prices-are-integer-cents,

after 20 commits on src/billing/** it is flagged for a check,

a newer decision that supersedes it takes its place, and kontext log --all keeps the chain.

Why not Claude Code's memory or CLAUDE.md?

	CLAUDE.md, AGENTS.md, `.claude/rules`	Claude Code auto memory	Memory services	kontext
Written by	people, as prose	the model, on its own	the model, from sessions	agents propose, people approve
Shared and reviewed	in pull requests	no: one user, one machine	per deployment, no review	in the pull request of the change
Reaches the agent	the root file every session; a nested `CLAUDE.md` when the agent reads files in its directory; a `.claude/rules` file with `paths:` when matching files are touched	the first 200 lines of its index, every session	by similarity to the query	ranked for the paths it names, within a budget
An entry records	paths at most (`.claude/rules`); no status, date or owner	when it was written	depends on the store	paths, status, date, author, last verification
When it goes stale	stays until someone notices	the model may rewrite it	the model may overwrite it	flagged after N commits on its paths
Works with	Claude Code reads `AGENTS.md` natively, as do Codex, Cursor and OpenCode; `.claude/rules` is Claude Code only	Claude Code	their plugin, SDK or MCP server	every MCP client and the CLI

CLAUDE.md, AGENTS.md, .claude/rules

Claude Code auto memory

Memory services

kontext

Written by

people, as prose

the model, on its own

the model, from sessions

agents propose, people approve

Shared and reviewed

in pull requests

no: one user, one machine

per deployment, no review

in the pull request of the change

Reaches the agent

the root file every session; a nested CLAUDE.md when the agent reads files in its directory; a .claude/rules file with paths: when matching files are touched

the first 200 lines of its index, every session

by similarity to the query

ranked for the paths it names, within a budget

An entry records

paths at most (.claude/rules); no status, date or owner

when it was written

depends on the store

paths, status, date, author, last verification

When it goes stale

stays until someone notices

the model may rewrite it

the model may overwrite it

flagged after N commits on its paths

Works with

Claude Code reads AGENTS.md natively, as do Codex, Cursor and OpenCode; .claude/rules is Claude Code only

Claude Code

their plugin, SDK or MCP server

every MCP client and the CLI

kontext does not replace CLAUDE.md or AGENTS.md: keep the few instructions every session needs there. Decisions are too many to load whole and too important to leave unreviewed. More on the alternatives

Under your control

Approval. Knowledge changes only through commits. With /.ai/ @acme/architects in CODEOWNERS and code-owner review required, no agent changes what every agent follows without that team's approval.

Audit trail. git log -- .ai and kontext log --all show who decided what, when, and what replaced it. With kontext's hooks installed, git log --grep "Decision: <id>" finds the commit that shipped a decision.

Policy in CI. kontext check fails a pull request on malformed entries, duplicate ids or high-confidence secrets.

No lock-in. The store is plain Markdown in your repository and stays readable without kontext.

kontextDeclarative team knowledge for coding agents