Self-maintaining code memory for AI coding agents: an Obsidian-compatible knowledge graph of your codebase, served over MCP and auto-curated from your work sessions.
Wellknown found it in public sources; nobody has proven control of it yet. Claiming takes one click if the repository is under your GitHub account, or a small file on your domain otherwise. Verified owners get the badge, 15-minute checks, status alerts, edits that outrank crawled data, and a ranking boost.
Agents can do it too: POST https://wellknown.network/api/v1/claims with {"agent":"tremula-mcp","method":"well_known_file"} — machine-readable steps at claim.json, guide at /docs/claim.
Everything here was measured by our prober or read from a registry. Nothing is self-reported.
Attributed to the source that supplied each field. Treated as claims, not facts.
# Tremula [](https://pypi.org/project/tremula-mcp/) [](https://pypi.org/project/tremula-mcp/) [](LICENSE) > *Populus tremuloides* — the quaking aspen. The Pando colony is thousands of > trunks sharing one living root system. Tremula gives your codebases the same > shape of memory: one knowledge graph per project, connected at the roots > where projects actually meet. **Tremula is self-maintaining memory for AI-assisted development.** It keeps an Obsidian-compatible markdown knowledge graph about your codebase — modules, key functions, architectural decisions, conventions, cross-service contracts — and maintains it automatically while you work. ## Why AI coding agents start every session amnesiac, and hand-written docs rot. Tremula closes both gaps with two cooperating loops: - **Reactive — an MCP server** exposing six tools to any MCP client: `search`, `get_context` (a topic *plus* its graph neighbors), `read_note`, `write_note`, `link_notes`, `split_note`. - **Ambient — lifecycle hooks** that work without being asked: capture session events to a cheap local log, inject the vault's index at session start, silently attach the few notes relevant to the files being touched, and distill durable knowledge (decisions, conventions, contracts) into notes in a debounced background process. Retrieval is a funnel, never a dump: a ~2k-token index at session start, graph-expanded context on demand, small working-context attachments per prompt. Context overhead stays constant no matter how large the vault grows. ## How the vault gets built and maintained | Mechanism | When | Cost | |---|---|---| | `tremula bootstrap --brief` | once, instant | zero LLM — docstrings + tree-sitter AST | | `tremula bootstrap [target ...]` | wherever you choose to go deep | ~one small LLM call per module | | …
Mapped onto the structured taxonomy from declared text and observed tool names. Confidence shown for derived entries.
Every source is kept verbatim. Field changes are logged as events.