AI coding agents forget everything between sessions. Switch from Claude Code to Cursor to Kiro to OpenCode and the context is gone -- decisions
Connect Ogham MCP to Claude, Cursor or any other MCP client and it stops being a tab you switch to. AI coding agents forget everything between sessions. Switch from Claude Code to Cursor to Kiro to OpenCode and the context is gone -- decisions, gotchas, the shape of your codebase -- so you repeat yourself, re-explain, and re-debug the. The ogham mcp mcp server is what makes that connection.
The toolset is worth reading before you wire it up, because it tells you what the integration is really for:
Retrieval — ** 97.2% R@10 on LongMemEval with one Postgres query (pgvector + tsvector CCF hybrid search). The paperEnd-to — end QA:** 85.8% on the AMB harness (500 questions, April 2026, strict substring judgeTyped — edge graph (v0.16)** -- store_triple / query_join for two-fact join queries against a controlled predicate vocabularyImporters — - Claude Code auto-memory, Claude.ai export, Linear issues, and JSONSkills — - ogham-research, ogham-recall, ogham-maintainSelf — hoster options** -- ONNX local embeddings (BGE-M3), optional FlashRank reranking, five embedding providersACT — R importance + Hebbian decay** -- recency, frequency, and surprise weighting (Anderson & Lebiere, 1998; [HebbRead — time fact extraction** -- verbatim storage with query-aware extraction at retrieval, so the ground truth stays re-extractable with differentAppend — only audit trail** -- every store, search, delete, and update logged to an audit_log table in the same Postgres instance, aligned with GDPR ArticleMethod — CommandSystem — AccuracyOgham — 97.2%Because this one is hosted, setup is mostly authentication — you point your client at the endpoint and approve access. Nothing runs on your machine, so there is no runtime to keep patched.
Configuration is passed through the environment: SUPABASE_URL, SUPABASE_KEY, EMBEDDING_PROVIDER, OPENAI_API_KEY, DATABASE_BACKEND, DATABASE_URL, OGHAM_HOST, MISTRAL_API_KEY. Treat anything key-shaped as a real credential — scope it to the minimum the server needs, and rotate it if it ever lands in a shared config.
This sits in the developer tooling group, where several servers overlap in what they claim to do but differ sharply once you actually set them up. Ogham MCP's toolset — Retrieval, End-to, Typed and 11 more — is a fair guide to whether it matches your workflow. It is maintained by ogham-mcp; worth a glance at recent repository activity before you build anything load-bearing on it.
This entry was verified against Ogham MCP's own documentation before publication; SyncDev keeps the directory reviewed rather than auto-generated.
| Tool | What it does |
|---|---|
| Retrieval | ** **97.2% R@10 on LongMemEval** with **one Postgres query** (pgvector + tsvector CCF hybrid search). The [paper](https://arxiv.org/abs/2410.10813) baseline is 78.4%. Other systems that report similar R@10 typically stac |
| End-to | end QA:** **85.8% on the [AMB harness](https://github.com/vectorize-io/agent-memory-benchmark)** (500 questions, April 2026, strict substring judge, GPT-5-mini reader; R@10 99.5%), and **0.554 nugget on [BEAM](https://ar |
| Typed | edge graph (v0.16)** -- store_triple / query_join for two-fact join queries against a controlled predicate vocabulary. [docs](https://ogham-mcp.dev/docs/typed-edges/) |
| Importers | - Claude Code auto-memory, Claude.ai export, Linear issues, and JSON. |
| Skills | - ogham-research, ogham-recall, ogham-maintain. |
| Self | hoster options** -- ONNX local embeddings (BGE-M3), optional FlashRank reranking, five embedding providers. |
| ACT | R importance + Hebbian decay** -- recency, frequency, and surprise weighting ([Anderson & Lebiere, 1998](https://act-r.psy.cmu.edu/about/); [Hebb, 1949](https://doi.org/10.4324/9781315735368)). Unaccessed memories fade; |
| Read | time fact extraction** -- verbatim storage with query-aware extraction at retrieval, so the ground truth stays re-extractable with different questions later ([Anthropic, arXiv:2510.05179](https://arxiv.org/abs/2510.05179 |
| Append | only audit trail** -- every store, search, delete, and update logged to an audit_log table in the same Postgres instance, aligned with GDPR Article 15 and [OTEL GenAI conventions](https://opentelemetry.io/docs/specs/semc |
| Method | Command |
| System | Accuracy |
| Ogham | **97.2%** |
| Category | R@10 |
| single-session-assistant | 100% |
{
"mcpServers": {
"ogham": {
"command": "uvx",
"args": ["ogham-mcp"],
"env": {
"SUPABASE_URL": "your-value",
"SUPABASE_KEY": "your-value",
"EMBEDDING_PROVIDER": "your-value",
"OPENAI_API_KEY": "your-value",
"DATABASE_BACKEND": "your-value",
"DATABASE_URL": "your-value",
"OGHAM_HOST": "your-value",
"MISTRAL_API_KEY": "your-value"
}
}
}
}Add to claude_desktop_config.json, then restart Claude Desktop.
| Variable | Description | Required |
|---|---|---|
| SUPABASE_URL | Endpoint or connection string the server talks to. | Yes |
| SUPABASE_KEY | Credential the server authenticates with. | Yes |
| EMBEDDING_PROVIDER | Configuration value read at startup. | Optional |
| OPENAI_API_KEY | Credential the server authenticates with. | Yes |
| DATABASE_BACKEND | Configuration value read at startup. | Optional |
| DATABASE_URL | Endpoint or connection string the server talks to. | Yes |
| OGHAM_HOST | Endpoint or connection string the server talks to. | Optional |
| MISTRAL_API_KEY | Credential the server authenticates with. | Yes |
Kill hallucinated APIs — version-accurate, up-to-date library documentation injected straight into context.
Microsoft's official browser automation server — drive a real browser through the accessibility tree, no screenshots needed.
GitHub's official server — repos, issues, pull requests, Actions and code security, straight from your assistant.
Issue tracking at the speed of conversation — Linear's official hosted server with OAuth and zero install.
Local repository surgery — status, diffs, commits, branches and history for any repo on disk.
Timezone sanity for AI — current time anywhere and correct conversions, without the model doing date math.