AgentIndex icon
AgentIndex
ToolsMCP ServersHOTCategoriesTrendingNewBest ListsCompareGuides
Submit Tool
ToolsMCP ServersHOTCategoriesTrendingNewBest ListsCompareGuides
Home/
Memory & Context/
cognee
cognee logo

cognee

Active·★ 33.4k·Apache-2.0·Updated 2026-11-18
★ Most Popular

Cognee is a persistent memory platform for agents that need to connect documents, code and previous conversations. Its value should be evaluated through useful retrieved evidence, especially where relationships across sources matter. A small set of stable project rules may be easier to maintain in an instruction file; graph memory becomes worth investigating as source volume and relationship complexity increase.

This review checked public documentation, package metadata and the remember implementation on November 4, 2026. PyPI listed Cognee 1.6.6 for Python 3.10 through 3.14. We did not install models, build a business graph or benchmark accuracy. The full article therefore provides a proposed evaluation plan, not measured retrieval or cost improvements, and identifies its example as unexecuted.

Distinguish session storage from permanent ingestion. A remember call with a session identifier first uses session memory, while the documented improvement bridge can move content into the permanent graph. Immediate recall in that same session is insufficient evidence of durable memory. Check processing completion, then query from another session and inspect the returned source. Treat local extraction, embeddings and generated answers as separate provider decisions.

Read the full review, verification scope and sources →

This write-up is based on the project's README and public repository data. We have not hand-tested the tool. Data as of 2026-11-12.

License
Apache-2.0
Latest release
v1.6.7 · 2026-11-06
Last commit
2026-11-12
Contributors
334
Forks
3,502
Open issues and PRs
1,121
First published
2023-08-16
PyPI package
cognee 1.6.7 · 45,220 downloads last week

Source: GitHub, npm and PyPI public APIs.

#agent-memory#agent-skills#ai#ai-agents#ai-memory#cognitive-architecture#cognitive-memory#context-engineering
$ Install
$ uv pip install "cognee[gliner]"
↗ Visit site★ GitHub
01

Features

01Provides remember, recall, improve and forget operations for stored knowledge.
02Uses a local GLiNER extractor and local embeddings when no LLM key is configured.
03Connects graph, vector and code context during retrieval.
04Offers Python, CLI, plugins and MCP integration paths.
02

Why choose it

+Connects documents, code and conversations in a reusable memory layer.
+Documents distinct ingestion, retrieval and deletion operations for evaluating the memory lifecycle.
+Offers a documented local text path with small extraction and embedding models.
03

Trade-offs

!Session recall alone does not verify that content reached permanent graph memory.
!Local models still need installation, downloads, storage and operational ownership.
!Retrieval quality, update behavior and isolation require tests against your own corpus.
05

Maintenance check

ActiveChecked: 2026-11-12
Last commit
2026-11-06
Commits, last 90 days
100+
Releases, last 90 days
23
Latest release
v1.6.7 (2026-11-06)
Open issues
1121
Recent issues with at least one comment
3/4
Commits per week, last 12 weeks
Commits per week, last 12 weeks: 114, 290, 227, 400, 178, 289, 337, 238, 374, 265, 174, 0

Oldest week on the left; the latest four weeks are orange. Source: GitHub commit activity.

Status rule: archived on GitHub = Archived; last commit within 90 days = Active; 91 to 365 days = Slowing; older = Stale. All figures come from the GitHub API on the date shown.

Most discussed open issues

  • #4778 Hackathon [Feature]: Add Rollbar data-source connector30 comments · 2026-08-28
  • #4811 Hackathon [Feature]: Add Reddit data-source connector29 comments · 2026-08-28
  • #3625 Hackathon: JSON Schema → Pydantic → cognee graph utilities — extend the base + full tests29 comments · 2026-06-28
  • #5776 Hackathon [Feature]: Add WooCommerce data-source connector28 comments · 2026-10-13
  • #5470 Hackathon [Feature]: Add Mattermost data-source connector28 comments · 2026-10-07

Issue titles are quoted from GitHub. Open the link to check whether a problem still applies to the version you use.

04

Who it fits

Pick it if
  • Your tasks depend on relationships across sources and sessions.
  • You can evaluate evidence quality and maintain the memory lifecycle.
Look elsewhere if
  • A maintained project file already contains all the context you need.
  • You need proven retrieval accuracy without running a corpus-specific pilot.
06

Quick start

1
$ uv pip install "cognee[gliner]"
07

Use cases

↳Retain project decisions, fixes and lessons across agent sessions.
↳Connect company documentation, conversations, tickets and code in shared memory.
↳Model domain entities and relationships for grounded context retrieval.
08

Alternatives

⚡Top VS:
VSmagic-context★2.5k→VSclaude-mem★118.0k→VSpensyve★101→VSobsidian-second-brain★5.4k→
magic-context logo
magic-context★ 2.5k
Unbounded context. Memory that manages itself. One session, for life. The hippocampus for coding agents, part of CortexKit.
VSCompare→
claude-mem logo
claude-mem★ 118.0k
Persistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More
VSCompare→
pensyve logo
pensyve★ 101
Universal memory runtime for AI agents
VSCompare→
obsidian-second-brain logo
obsidian-second-brain★ 5.4k
Persistent memory for Claude Code, Codex and 6 other AI agents, stored as plain markdown in your Obsidian vault. New: Crew, one shared logbook so Claude Code, Codex, Cursor and Grok Bot know what each other did. 48 commands: semantic search, self-rewriting notes, key-less research, scheduled agents.
VSCompare→
origin logo
origin★ 498
Local-first Rust daemon for AI work. Git-versioned memories, distilled wiki pages, sessions for Claude Code, Cursor, Codex, and any MCP client.
VSCompare→
Ori-Mnemos logo
Ori-Mnemos★ 337
Local-first persistent agentic memory powered by Recursive Memory Harness (RMH). Open source must win.
dreamgraph logo
dreamgraph★ 128
A semantic cognitive engine for software systems. By combining MCP context with iterative "dream cycles," DreamGraph builds an abstract representation of system intent—acting as a universal architect capable of autonomous reasoning, system abstraction, and cross-ecosystem transcompilation.
VSCompare→
ctx-open logo
ctx-open★ 60
CTX is the standard Cognitive Version Control System for AI. Structured working memory for AI agents. CTX preserves goals, tasks, evidence, decisions, conclusions, runbooks, and origins so agents can continue work instead of restarting from scratch.
VSCompare→
See all alternatives →

Comments

Log in to leave a comment

No comments yet. Be the first!

On this page
01Features02Why choose it03Trade-offs05Maintenance check04Who it fits06Quick start07Use cases08Alternatives
topoteretes /
cognee
↗

Graph-based memory for agents that need evidence across documents and sessions

33.4k3.5kApache-2.0
Last commitToday
StatusActive
CategoryMemory & Context
Contributors334
Latest releasev1.6.7
Links
Documentation↗Discussion↗Issues↗Releases↗
Cloud Infrastructure
Ad

Deploy on DigitalOcean

Host your MCP servers, Docker containers, and AI agents with $200 free credit over 60 days.

Claim $200 Free Credit
© 2026 AgentIndex.app|Built by a 10-year iOS Developer.
Best by use caseAboutContactAdvertisePrivacy PolicyTerms of UseQYSGitHubBuy me a coffee ☕

Browse by Category

MCP ServersCode AssistantWorkflow AutomationRAG / Knowledge BaseMulti-AgentBrowser AutomationLLM InfraDev ToolingObservability

Not affiliated with Anthropic, OpenAI or Microsoft.