MindGraph Adds Persistent Graph Memory for Coding Agents

TL;DR: Multiple posts discuss MindGraph, a state management foundation for coding agents. It aims to solve the "amnesia" and context loss issues in tools like Claude Code and Codex during session ends, tool switching, or multi-agent concurrency. The core approach persists tasks, iterations, blockers, and lessons learned as a graph state, using symbol-based code anchors instead of line numbers to link code context. For long-term, multi-player, or multi-tool AI programming workflows, such a foundation is critical for continuous agent operation.

Confirmed

ShanRizvi introduced MindGraph for Claude Code and Codex, positioning it as a persistent graph memory across sessions for coding agents.
The system writes goals, plans, failed attempts, reasons, blockers, and lessons into a graph-managed working state; new sessions only need a brief context injection to resume previous work.
Its memory persistence uses a code anchor mechanism: establishing stable identities based on code symbols rather than fragile line numbers, ensuring associated experiences survive code refactoring.
The system queries a local code graph index to retrieve relevant context on demand, rather than stuffing the entire history into the prompt.
In multi-agent scenarios, MindGraph handles concurrency using leases and Epoch guards. If two agents handle the same task, the lagging one receives a clear expiration or conflict signal, preventing old states from overwriting new ones.
Performance tests indicate the write overhead of this graph state approach is lower than a single LLM round-trip, making it viable for daily development workflows.

Unconfirmed

Direct support is confirmed for Claude Code and Codex. A repost mentioned Supermemory attempting to connect Claude Code, Cursor, Opencode, Gemini, etc., but this is not a directly confirmed support scope for MindGraph in these posts.

Why It Matters

@Asleep-Pilot-4142 noted that switching between Cursor, Claude Code, and Codex often causes context loss, as each tool is unaware of the other's actions, forcing manual syncs of key decisions.
If solutions like MindGraph succeed, their significance goes beyond "remembering chat history" by persisting task states, code locations, failure experiences, and concurrency control together. This evolves agents from one-off conversational tools into sustainable, collaborative development executors.

2026-07-28 ~ 2026-07-29 · 8 related posts

Primary sources

MindGraph gives coding agents durable memory that survives sessions — ShanRizvi · 2026-07-29 19:47 (UTC+08:00)
Coding Agent Memory: Using Symbol Anchors Instead of Line Numbers — ShanRizvi · 2026-07-29 22:20 (UTC+08:00)
Solving Multi-Agent Concurrency: Leases with Epoch Fencing — ShanRizvi · 2026-07-29 22:20 (UTC+08:00)

Supermemory pitches shared context transfer across Claude Code, Cursor, and Gemini — builditwithjoe · 2026-07-28
How one developer is trying to keep Cursor, Claude Code, and Codex sharing the same context — Asleep-Pilot-4142 · 2026-07-29
MindGraph adds durable graph memory to Claude Code and Codex sessions — ShanRizvi · 2026-07-29
[source] MindGraph gives coding agents durable memory that survives sessions — ShanRizvi · 2026-07-29
Fixing Coding Agent Amnesia: Graph-Managed State Substrate Benchmarks — ShanRizvi · 2026-07-29
[source] Coding Agent Memory: Using Symbol Anchors Instead of Line Numbers — ShanRizvi · 2026-07-29
[source] Solving Multi-Agent Concurrency: Leases with Epoch Fencing — ShanRizvi · 2026-07-29
Graph Memory and Lease Isolation: Multi-Agent State Management in Practice — ShanRizvi · 2026-07-29