REF. C: COMPARATIVE ANALYSIS

How kontexta compares

MATRIX C: CAPABILITY COMPARISON

Capability	CLAUDE.md	Vendor memory	mem0	Linksee	Cosmos	kontexta
Setup cost	None, just a file	None, built in	SDK integration in your app	npx linksee-memory setup	Curl install + macOS handler for iMessage / calendar	npx kontexta start + kontexta.json
Cross-agent portability	Per-agent flavored files drift apart	Locked to one vendor	App-level, not agent-level	One SQLite read by Claude, Cursor, Codex, Gemini	One .polarity graph across any MCP client	Same MCP surface for Claude Code, Cursor, Codex, Gemini, Aider
Retrieval model	Whole file dumped into context	Whole file / vendor-managed	Vector + graph (semantic)	6-layer structured memory + AST diff cache	Knowledge graph, NL-synthesized on query	Deterministic FTS5 + regex; surgical section reads
Self-editing memory	Manual edits only	Append-only per session	Append-only + dedup	Drift detection vs locked decisions (map.yaml)	Append-only, graph re-synthesized on read	Distill loop rewrites summaries from event log
Command execution	N/A	Vendor-defined tools	N/A (memory only)	N/A (memory only)	N/A (memory + graph only)	Sandboxed Hands with per-command contracts and approval
Token accounting	None	None	None exposed to agent	Diff cache reduces re-read cost	None exposed to agent	Every response carries est_tokens / size_bytes
Storage	Repo file (clutters PRs)	Vendor cloud	Self-host or hosted, vector DB	Local SQLite	Hosted cloud graph (.polarity export)	Local SQLite, git-synced markdown vault
Data ingestion	None	Chat history only	Via SDK ingestion	Agent-writes only	iMessage, browser, calendar, shell history	URL clipping + agent-writes; no personal-data vacuum
Best at	Static project conventions	Zero-config personal memory	Semantic recall inside one app	Catching agent drift from locked decisions	Personal knowledge graph across life apps	Multi-agent handoff + governed local execution

REF. C.1: DESIGN RATIONALE

Why kontexta

R-01

One vault, every agent

Same MCP surface for Claude Code, Cursor, Codex, Gemini, Aider, Copilot. Switch tools mid-project; the vault and journal follow.

R-02

Self-editing memory

Everyone else appends. Kontexta's distill loop rewrites the raw event log into per-topic markdown summaries the next session searches instead of re-deriving.

R-03

Governed execution

Every Hands command is declared in kontexta.json. Approval gates for the destructive ones. No rogue shell.

R-04

Deterministic retrieval

SQLite FTS5, not vectors. Exact keyword and regex search, no embedding hallucination, no third-party API.

R-05

Context economy

Surgical fetching: outlines, sections, search excerpts. Every response carries est_tokens so agents budget intentionally.

R-06

Local-first architecture

Your vault on your SSD. No cloud sync, no upload. KONTEXTA_DATA_DIR points wherever you want.