AgentIndex icon
AgentIndex
ToolsMCP ServersHOTCategoriesTrendingNewBest ListsCompareGuides
Submit Tool
ToolsMCP ServersHOTCategoriesTrendingNewBest ListsCompareGuides
Home/
Memory & Context/
aoci-code
aoci-code logo

aoci-code

Active·★ 2.9k·NOASSERTION·Updated 2026-11-21
★ Essential

AOCI-CODE keeps a persistent, Git-versioned index of your whole codebase inside the repository, written by your coding agent and governed by a local MCP server. The idea in the README is that the agent reads the index once and knows the system, instead of searching and re-reading files for every task. Each file gets one line: what it is responsible for (F), what must be read with it (R), what callers depend on (A) and what cannot be inferred from code but must not be gotten wrong (S), plus a short tag for layer, domain, importance and size. A few hundred lines are meant to cover a whole system.

The first index is built by your agent file by file, in resumable batches, and the README estimates about an hour per 200,000 lines, depending on the model. It claims a project of about 500,000 lines can be indexed, and says a 700,000-line commercial system is developed this way with an index of about 300K tokens. After that, the MCP server detects code changes and issues entries to update, and the agent fills them in as it finishes tasks. MySQL and PostgreSQL schemas (and constrained openGauss 6.0.5) can be indexed too, using table structure only.

The README says it is local only: it reads source code and database table structure, never business data, uploads nothing, opens no Internet connection, and keeps database credentials as environment-variable names. Setup is done by pasting an instruction into your agent that downloads a release, runs `init` and `scan`, then restarting the agent and asking it to build the index; an `aoci ui` panel shows status. The license is FSL-1.1-MIT (Functional Source License), which is source-available rather than open source. Last push was 2026-10-13. This review is based on the README and repository data; we have not run it ourselves.

This write-up is based on the project's README and public repository data. We have not hand-tested the tool. Data as of 2026-11-19.

License
Not detected by GitHub
Latest release
v0.1.1 · 2026-11-04
Last commit
2026-11-17
Contributors
16
Forks
231
Open issues and PRs
18
First published
2026-08-08

Source: GitHub, npm and PyPI public APIs.

#Codebase Mapping#AI Agent Cognition#Git Versioning#Context Management#Software Governance#Database Integration#Developer Tooling#Local-first Architecture
$ Install
$ git clone https://github.com/aoci-spec/aoci-code.git && cd aoci-code && make build
↗ Visit site★ GitHub
01

Features

01Provides a persistent, Git-versioned codebase map for AI agents.
02Enables rapid onboarding for agents to existing large codebases (up to 500k lines).
03Offers automatic, incremental index maintenance after the initial build.
04Integrates both code and database schema knowledge for comprehensive system understanding.
05Operates locally and securely, with no internet upload or stored credentials.
02

Why choose it

+A compact per-file index (responsibility, dependencies, public surface, must-not-break notes) that lives in the repo and is versioned by Git
+Meant to survive a change of agent, person or conversation: one read of the index picks up where work left off
+The MCP server detects code changes and tells the agent which entries to update, so maintenance is incremental
+Local only per the README: reads source and table structure, uploads nothing; works with Codex, Claude Code, Cursor and OpenCode
03

Trade-offs

!The first index is slow (the README estimates about an hour per 200,000 lines) and uses your agent's model tokens
!The index is only as good as the model that wrote it, and the 500K to 700K line figures are the project's own claims
!FSL-1.1-MIT is a source-available license with usage restrictions, not an OSI open-source license, so read the LICENSE
!The setup relies on pasting a download-and-install instruction to your agent, so verify the release package it downloads
05

Maintenance check

ActiveChecked: 2026-11-06
Last commit
2026-11-06
Commits, last 90 days
100+
Releases, last 90 days
26
Latest release
v0.2.0-rc1 (2026-11-05)
Open issues
11
Recent issues with at least one comment
19/21
Commits per week, last 12 weeks
Commits per week, last 12 weeks: 14, 9, 24, 19, 30, 19, 7, 19, 59, 0, 98, 52

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

  • #54 支持多 git 仓库容器目录、灵活选择索引范围与渐进式建立索引12 comments · 2026-09-15
  • #84 Windows Defender aoci.exe 下载后移除 - 检测到病毒,导致无法安装4 comments · 2026-09-26
  • #135 Feature: index MongoDB data models — extend `database source` (currently PostgreSQL/MySQL/openGauss) to MongoDB2 comments · 2026-10-17
  • #34 Port the product-code parts of #28 after v0.1.02 comments · 2026-09-07
  • #155 [Feature] 支持将索引与运行状态集中到单个 .aoci 目录,并提供安全迁移流程1 comments · 2026-10-27

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 codebase or database schema is large and complex, causing coding agents to burn tokens and lose context across sessions.
  • You want a local-first, Go-based, Git-versioned architectural knowledge map that coding agents read before touching any files.
Look elsewhere if
  • Your project is a simple small repo or script where agents can easily fit the whole tree in context without an index map.
  • You need general conversational memory for chat logs or personal tasks rather than software architecture topology (use Mem0 or mcp-memory-service).
06

How it differs

codebase-memory-mcp logovs codebase-memory-mcp★ 49.0k

aoci-code maintains a Git-versioned structural map of symbols and schemas; codebase-memory-mcp indexes codebases into vector embeddings for semantic search.

Full comparison →
code-graph-rag logovs code-graph-rag★ 5.4k

aoci-code emphasizes Git-governed agent architecture contracts; code-graph-rag focuses on AST graph-database construction and graph traversal.

Full comparison →
07

Compatibility

Go
Language
Verified via docs
Linux
OS
Verified via docs
MacOS
OS
Verified via docs
Windows
OS
Verified via docs
Codex
AI Agent Host
Verified via docs
Claude Code
AI Agent Host
Verified via docs
08

Quick start

1
$ git clone https://github.com/aoci-spec/aoci-code.git
2
$ cd aoci-code
3
$ make build
09

Use cases

↳Developing large software systems with AI agents, ensuring agents retain context and avoid re-learning on every task.
↳Onboarding new AI agents or developers to existing complex codebases quickly and efficiently.
↳Maintaining project continuity and shared understanding across different agents, conversations, or team changes.
↳Enhancing AI agent's ability to perform complex development tasks (frontend, backend, database, tests) with a full system overview.
10

Alternatives

⚡Top VS:
VScodebase-memory-mcp★49.0k→VScode-graph-rag★5.4k→VSdeja-vu★1.6k→
codebase-memory-mcp logo
codebase-memory-mcp★ 49.0k
High-performance code intelligence MCP server. Indexes codebases into a persistent knowledge graph, with sub-millisecond queries across 150+ languages. Single static binary, zero dependencies.
VSCompare→
code-graph-rag logo
code-graph-rag★ 5.4k
The ultimate RAG for your monorepo. Query, understand, and edit multi-language codebases with the power of AI and knowledge graphs
VSCompare→
projectmem logo
projectmem★ 877
Local-first AI coding memory for AI agents. Records issues, attempts, fixes and decisions, then warns your agent before it repeats an approach that already failed. Native MCP server for Claude Code, Cursor, Antigravity and Codex. 100% local, no cloud, no telemetry. MIT.
GoodMemory logo
GoodMemory★ 21
Local-first, auditable memory layer for AI apps and coding agents — Codex, Claude Code, MCP, HTTP, TypeScript, and Python.
deja-vu logo
deja-vu★ 1.6k
Your agents already solved this. deja finds it — retroactive search and MCP recall over the session logs your coding agents already wrote. Cross-agent, SSH-portable, one zero-dep binary, fully local.
VSCompare→
okf-agent-memory logo
okf-agent-memory★ 870
Git-native persistent memory for AI coding agents. Implements Google OKF v0.2 with sub-300µs in-memory BM25 search, embedded MCP server, and progressive disclosure. Slashes token bloat by 80% with zero external databases or dependencies. Built in pure Go.
agent-memory logo
agent-memory★ 15
Local, searchable project memory for AI coding agents. Markdown source of truth, MCP interface, safe structured updates — no cloud.
AgentRecall-X logo
AgentRecall-X★ 376
Correction-first persistent memory for AI agents. MCP server + SDK + CLI. Compounds across sessions.
See all alternatives →

Featured in these lists

  • Best Memory Tools for AI Agents and Coding Assistants

Comments

Log in to leave a comment

No comments yet. Be the first!

On this page
01Features02Why choose it03Trade-offs05Maintenance check04Who it fits06How it differs07Compatibility08Quick start09Use cases10Alternatives
aoci-spec /
aoci-code
↗

Git-versioned map of your codebase written by your coding agent and served by a local MCP server; FSL-licensed, source-available

2.9k231
Last commit1d ago
StatusActive
CategoryMemory & Context
Contributors16
Latest releasev0.1.1
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.