OpenContext claims its project-local MCP server gives AI coding agents private, durable memory across sessions, potentially making persistent context portable across compatible coding tools.

state: resolvedheat: lowuncertainty: mediumknownscott: lowagent-memory coding-agents mcpOpenContext

What is this?

OpenContext is a lightweight personal context and knowledge store maintained in the `0xranx/OpenContext` GitHub repository for AI assistants and coding tools including Cursor, Claude Code, and Codex. It claims to give agents persistent memory across sessions through a project-local MCP server, keeping context private and potentially reusable by MCP-compatible clients. The supplied snippets describe these capabilities but do not independently establish the privacy guarantees, portability limits, implementation details, or comparative performance.

Why it matters to Scott

The position is already held in Scott’s “Markdown OS” and “Long-Running Agents” frameworks and implemented in projects such as Search Conversations: durable state should live outside individual coding sessions in portable, inspectable infrastructure. OpenContext is another closely tracked local-memory implementation, but the supplied evidence offers no verified architectural, portability, or performance advance that would change Scott’s view or builds.

ip:framework.markdown-osip:framework.long-running-agentsdev:project.search-conversationsdev:project.mcp-ip-wikiradar:concept.agent-memoryradar:concept.mcpradar:awareness-local-agent-memoryradar:rune-persistent-coding-context

queries asked of Scott's wikis

project-local memory for coding agents
MCP as portable agent infrastructure
durable context across coding sessions
agent-maintained project knowledge stores
private local-first agent memory
memory retrieval and context compression

Measured heat

no measured readings yet — the hourly heat pass fills this in

How the heat travelled

no chain yet — the hourly chain pass fills this in

Evidence (10) — ⭐ canonical anchor

source	object	author	score	comments
🟧 hn	OpenContext – Persistent, project-local memory for AI coding agents via MCP	slxca	3	0
🟧 echo.blog ⭐	OpenContext presents persistent, project-local memory for AI coding agents through MCP.	OpenContext	—	—
🟧 hn	Show HN: Memnest, local-first memory shared by pi, Claude Code and Codex	blue-b	1	0
🟧 hn	Show HN: Rta-Smriti – local-first project memory for coding agents	SulabhDubey	1	0
🟠 reddit	my coding agent keeps re-proposing the thing we killed three months ago ClaudeAI	thePangee	0	16
🟧 hn	Aplexica – keep memory and conversations when you leave Cursor for Claude	yruzin	1	0
🟧 hn	End AI agent lock-in. Start conversation with one agent – continue with another	yruzin	1	0
🟧 hn	ThoughtDAG can now turn local coding-agent sessions into editable context graphs	chatchan	1	1
🟠 reddit	Auto-memory is per user. The decision is per repo. I keep mixing those up. ClaudeAI	architdhamija6	1	5
🟧 hn	Show HN: DeepMem plugin for Claude Code – persistent semantic memory via MCP	deepmem	2	0

Interpretation history

2026-09-04T16:34:51Z
DeepMem is another repetitive implementation of MCP-based persistent coding-agent memory, making the broader project-local memory pattern established enough to absorb. It still provides no substantive validation of OpenContext’s specific privacy, cross-client portability, retrieval quality, or interoperability claims.
2026-09-04T15:22:59Z
evidence attached: hn.story.49565805 — This independent MCP plugin is concrete corroborating evidence that project-local semantic memory is becoming portable across Claude Code sessions.
2026-09-04T09:25:36Z
The new user report sharpens the architecture into two distinct layers: private per-user agent memory and repository-shared decision memory. It strengthens the broader design requirement but adds no validation of OpenContext’s privacy, portability, interoperability, or implementation quality.
2026-09-04T09:22:14Z
evidence attached: reddit.post.1w6yfdh — Concrete user experience supports the open case that durable project-local memory must be separated from private per-user agent memory.
2026-09-02T18:34:38Z
ThoughtDAG broadens the established pattern from durable stores to editable context graphs, but title-only evidence offers no validation of architecture, interoperability, retrieval quality, or OpenContext’s specific claims. This remains repetitive implementation evidence rather than acceleration or a meaningful change for Scott.
2026-09-02T18:22:50Z
evidence attached: hn.story.49539486 — Turning coding-agent sessions into editable context graphs is direct evidence about durable, inspectable project memory.
2026-09-02T16:51:30Z
The Aplexica repost and refreshed discussion repeat the established demand for portable, durable coding-agent context without adding technical evidence about interoperability, privacy, migration fidelity, or quality. The broader pattern is well corroborated but not accelerating, and OpenContext’s specific claims remain unvalidated.
2026-09-02T16:22:56Z
evidence attached: hn.story.49537958 — shared external link with case evidence
2026-09-01T17:39:10Z
Aplexica adds another implementation aimed at carrying conversations and memory between coding tools, strengthening the broader portability pattern. Its title-only evidence still does not validate OpenContext’s privacy, interoperability, migration fidelity, or implementation quality, so this is repetitive corroboration rather than acceleration.
2026-09-01T17:26:48Z
evidence attached: hn.story.49524480 — Crossing from Cursor to Claude while preserving conversations is direct supporting evidence for portable, durable coding-agent context.
2026-08-31T21:49:10Z
The refreshed discussion adds an important lifecycle constraint: stored architectural decisions need dates and status because rejected options can become viable again. This refines the established design pattern but provides no new validation of OpenContext’s privacy, portability, interoperability, or quality claims.
2026-08-31T19:43:10Z
The discussion sharpens the recurring need from generic persistence to preserving rejected alternatives and their rationale, but this remains aligned with Scott’s existing ADR/Markdown-OS approach. It adds no validation of OpenContext’s privacy, portability, interoperability, or implementation quality.
2026-08-31T17:24:46Z
evidence attached: reddit.post.1w3j7tl — This user report independently illustrates coding agents repeatedly losing architectural rationale, directly supporting the need for durable project-local memory.
2026-08-31T16:38:35Z
A third independent implementation further establishes project-local coding-agent memory as a recurring pattern, but still does not validate OpenContext’s specific privacy, portability, interoperability, or quality claims. The added evidence is repetitive corroboration rather than a meaningful advance for Scott.
2026-08-31T16:24:40Z
evidence attached: hn.story.49511544 — This is independent corroboration of the emerging pattern of private, project-local memory preserving coding-agent context across sessions.
2026-08-30T03:23:34Z
Memnest provides a second implementation line for local, cross-tool coding-agent memory, corroborating the broader pattern but not OpenContext’s specific privacy, portability, or quality claims. This remains an already-known, low-relevance implementation pattern for Scott rather than a demonstrated advance.
2026-08-30T03:22:34Z
evidence attached: hn.story.49495320 — The released Memnest repository independently supports the episode of private, local, cross-tool persistent memory for coding agents.
2026-08-30T00:27:15Z
No substantive evidence has arrived beyond the original product claim; a trivial engagement increase without discussion does not validate privacy, portability, or implementation quality. The case remains a low-relevance example of a pattern Scott already tracks.
2026-08-30T00:26:15Z
grounded: known/low — The position is already held in Scott’s “Markdown OS” and “Long-Running Agents” frameworks and implemented in projects such as Search Conversations: durable sta
2026-08-30T00:24:17Z
case created — A usable first-party artifact introduces a bounded implementation of private cross-session coding-agent memory in a hot engineering area.

Decision trace

09-05 02:34resolveDeepMem is another repetitive implementation of MCP-based persistent coding-agent memory, making the broader project-local memory pattern established enough to absorb. It still provides no substantive
09-05 02:34alert_silentThe new delta is a title-only, low-engagement implementation with no technical results or consequential access change; it reinforces an already-established pattern and can wait for routine coverage.
09-05 02:34alert_routeThe new delta is a title-only, low-engagement implementation with no technical results or consequential access change; it reinforces an already-established pattern and can wait for routine coverage.
09-05 02:21sensor_dirtycomment_update
09-05 01:23alert_silentDeepMem appears to be another persistent semantic-memory implementation for Claude Code via MCP, but the supplied evidence is only a low-engagement Show HN title with no architectural, portability, re
09-05 01:23alert_routeDeepMem appears to be another persistent semantic-memory implementation for Claude Code via MCP, but the supplied evidence is only a low-engagement Show HN title with no architectural, portability, re
09-05 01:22attachThis independent MCP plugin is concrete corroborating evidence that project-local semantic memory is becoming portable across Claude Code sessions.
09-05 01:22propose_attachThis independent MCP plugin is concrete corroborating evidence that project-local semantic memory is becoming portable across Claude Code sessions.
09-04 19:25repriceThe new user report sharpens the architecture into two distinct layers: private per-user agent memory and repository-shared decision memory. It strengthens the broader design requirement but adds no v
09-04 19:25alert_silentThis is a useful design distinction already aligned with Scott’s Markdown OS approach, not a new release, demonstrated capability, adoption event, or architectural breakthrough; the next briefing is s
09-04 19:25alert_routeThis is a useful design distinction already aligned with Scott’s Markdown OS approach, not a new release, demonstrated capability, adoption event, or architectural breakthrough; the next briefing is s
09-04 19:23alert_silentThe post usefully distinguishes user-specific auto-memory from repository-shared decisions, but it reports no new product capability, architecture, benchmark, or access change. The lesson already matc
09-04 19:23alert_routeThe post usefully distinguishes user-specific auto-memory from repository-shared decisions, but it reports no new product capability, architecture, benchmark, or access change. The lesson already matc
09-04 19:22attachConcrete user experience supports the open case that durable project-local memory must be separated from private per-user agent memory.
09-04 19:21propose_attachConcrete user experience supports the open case that durable project-local memory must be separated from private per-user agent memory.
09-03 04:34repriceThoughtDAG broadens the established pattern from durable stores to editable context graphs, but title-only evidence offers no validation of architecture, interoperability, retrieval quality, or OpenCo
09-03 04:34alert_silentThe new delta is another unverified implementation claim within a pattern Scott already tracks closely; no release detail, demonstrated capability, adoption, or architectural advance makes waiting for
09-03 04:34alert_routeThe new delta is another unverified implementation claim within a pattern Scott already tracks closely; no release detail, demonstrated capability, adoption, or architectural advance makes waiting for
09-03 04:23alert_silentThoughtDAG is another low-evidence implementation of durable, tool-independent coding-agent context, already covered by Scott’s existing Markdown OS and long-running-agent frameworks. The supplied HN
09-03 04:23alert_routeThoughtDAG is another low-evidence implementation of durable, tool-independent coding-agent context, already covered by Scott’s existing Markdown OS and long-running-agent frameworks. The supplied HN
09-03 04:22attachTurning coding-agent sessions into editable context graphs is direct evidence about durable, inspectable project memory.
09-03 04:22propose_attachTurning coding-agent sessions into editable context graphs is direct evidence about durable, inspectable project memory.
09-03 02:51repriceThe Aplexica repost and refreshed discussion repeat the established demand for portable, durable coding-agent context without adding technical evidence about interoperability, privacy, migration fidel
09-03 02:51alert_silentThe new delta is repetitive amplification of an implementation pattern Scott already tracks, with no verified release change, adoption signal, or architectural advance that would make the next briefin
09-03 02:51alert_routeThe new delta is repetitive amplification of an implementation pattern Scott already tracks, with no verified release change, adoption signal, or architectural advance that would make the next briefin
09-03 02:23alert_silentThis is a duplicate Aplexica repost asserting cross-agent conversation portability, with no release details, architecture, compatibility evidence, or verified advance beyond the already tracked projec
09-03 02:23alert_routeThis is a duplicate Aplexica repost asserting cross-agent conversation portability, with no release details, architecture, compatibility evidence, or verified advance beyond the already tracked projec
09-03 02:22attachshared external link with case evidence
09-03 02:21propose_attachshared external link with case evidence
09-02 03:39repriceAplexica adds another implementation aimed at carrying conversations and memory between coding tools, strengthening the broader portability pattern. Its title-only evidence still does not validate Ope
09-02 03:39alert_silentThe new artifact reinforces an implementation pattern Scott already tracks but supplies no verified technical advance, adoption signal, or consequential access change; it can wait for the next briefin
09-02 03:39alert_routeThe new artifact reinforces an implementation pattern Scott already tracks but supplies no verified technical advance, adoption signal, or consequential access change; it can wait for the next briefin
09-02 03:27alert_silentAplexica appears to be another cross-tool, project-memory implementation, but the supplied evidence is only a low-engagement repository link and title. It provides no verified architectural, portabili
09-02 03:27alert_routeAplexica appears to be another cross-tool, project-memory implementation, but the supplied evidence is only a low-engagement repository link and title. It provides no verified architectural, portabili
09-02 03:26attachCrossing from Cursor to Claude while preserving conversations is direct supporting evidence for portable, durable coding-agent context.
09-02 03:24propose_attachCrossing from Cursor to Claude while preserving conversations is direct supporting evidence for portable, durable coding-agent context.
09-01 07:49repriceThe refreshed discussion adds an important lifecycle constraint: stored architectural decisions need dates and status because rejected options can become viable again. This refines the established des
09-01 07:49alert_silentThe new commentary is a useful implementation nuance, not a consequential product, adoption, or capability delta; it can wait for routine review.
09-01 07:49alert_routeThe new commentary is a useful implementation nuance, not a consequential product, adoption, or capability delta; it can wait for routine review.
09-01 07:21sensor_dirtycomment_update