mindtrail 0.1.0__tar.gz

mindtrail-0.1.0/.gitignore

@@ -0,0 +1,17 @@
+.venv/
+__pycache__/
+*.py[cod]
+*.egg-info/
+.pytest_cache/
+.mypy_cache/
+.ruff_cache/
+dist/
+build/
+*.db
+*.db-wal
+*.db-shm
+.env
+
+# Planning documents (kept local)
+IMPLEMENTATION_STATUS.md
+docs/design/

mindtrail-0.1.0/CHANGELOG.md

@@ -0,0 +1,51 @@
+# Changelog
+
+All notable changes are documented here. The format follows
+[Keep a Changelog](https://keepachangelog.com/en/1.1.0/) and the project uses
+[Semantic Versioning](https://semver.org/).
+
+## [Unreleased]
+
+## [0.1.0] - 2026-10-09
+
+First public release on PyPI and the MCP Registry.
+
+### Changed
+- `mindtrail[semantic]` now uses `BAAI/bge-base-en-v1.5` (210 MB) by default. It is ahead on
+  every bundled set (holdout v2: 94% recall@5, 100% abstention). Choose another model with
+  `MINDTRAIL_EMBEDDING_MODEL`, for example `BAAI/bge-small-en-v1.5` (67 MB). Existing memories
+  are re-embedded automatically when the server starts.
+- Cross-encoder reranking is off by default; enable it with `MINDTRAIL_RERANKER=<model id>`.
+- Server instructions now tell agents to call `recall` before any project-specific answer,
+  even a one-line one. In the agent evaluation this fixed the cases where the agent answered
+  from habit (`npm install`, `str | None` on a Python 3.9 project) instead of checking memory.
+- Renamed the project from CogMem to **Mindtrail** (package, CLI, `MINDTRAIL_*` variables,
+  `~/.mindtrail`).
+- The `local-embeddings` extra is now `semantic`. Neural models are stored in
+  `~/.mindtrail/models`, and `mindtrail init` pre-downloads them.
+- Retrieval now drops weak keyword matches (below 40% of the best BM25 score, or without vector
+  support) and uses calibrated similarity floors per embedder, so a query with no stored answer
+  is far more likely to return nothing.
+
+### Fixed
+- Clones of the same repository in differently named folders now share one project space.
+  The space name came from the folder instead of the git remote. In a repository whose folder
+  name differs from its remote's repo name, memories stored before this fix stay under the
+  old space id.
+
+### Added
+- Agent-in-the-loop evaluation (`benchmarks/agent_eval.py`): real Claude Code sessions,
+  with and without Mindtrail.
+- `holdout-v2` retrieval set and LoCoMo (`mindtrail bench --dataset locomo:test`), with
+  bootstrap confidence intervals in every report.
+- Retrieval benchmark (`mindtrail bench`) with bundled dev and held-out datasets: recall@k,
+  MRR, nDCG, abstention, stale/foreign leak rate and latency. CI gates on the results.
+- Core memory engine: validated immutable memory records, SQLite storage with FTS5, hybrid
+  keyword + vector retrieval, token-budgeted context, versioned updates, supersession,
+  validity windows, hard/soft deletion and secret filtering on write.
+- Offline hashing embedder (default) and optional local neural embeddings via fastembed.
+- MCP server over stdio with a three-tool core profile (`remember`, `recall`, `forget`) and a
+  `full` profile that adds `search_memory`, `get_context`, `update_memory` and `get_memory`.
+- Automatic project scoping from the git remote, shared across tools working in the same repo.
+- `mindtrail` CLI: `serve`, `init`, `doctor`, `remember`, `recall`, `forget`, `list`, `stats`,
+  `export` and `reindex`.

mindtrail-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Mindtrail contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

mindtrail-0.1.0/PKG-INFO

@@ -0,0 +1,233 @@
+Metadata-Version: 2.5
+Name: mindtrail
+Version: 0.1.0
+Summary: Persistent, local-first memory for AI coding agents over MCP.
+Project-URL: Homepage, https://github.com/RohitDeshmukh-1/Mindtrail-MCP
+Project-URL: Documentation, https://github.com/RohitDeshmukh-1/Mindtrail-MCP/tree/main/docs
+Project-URL: Issues, https://github.com/RohitDeshmukh-1/Mindtrail-MCP/issues
+Project-URL: Changelog, https://github.com/RohitDeshmukh-1/Mindtrail-MCP/blob/main/CHANGELOG.md
+Author: Mindtrail contributors
+License-Expression: MIT
+License-File: LICENSE
+Keywords: ai-agents,claude-code,cursor,llm,mcp,memory,rag
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Developers
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Topic :: Software Development :: Libraries
+Classifier: Typing :: Typed
+Requires-Python: >=3.11
+Requires-Dist: mcp<3,>=2.2
+Requires-Dist: numpy>=1.26
+Requires-Dist: pydantic<3,>=2.7
+Provides-Extra: dev
+Requires-Dist: mypy>=1.11; extra == 'dev'
+Requires-Dist: pre-commit>=3.7; extra == 'dev'
+Requires-Dist: pytest>=8; extra == 'dev'
+Requires-Dist: ruff>=0.6; extra == 'dev'
+Provides-Extra: semantic
+Requires-Dist: fastembed>=0.4; extra == 'semantic'
+Description-Content-Type: text/markdown
+
+<div align="center">
+
+# 🧭 Mindtrail
+
+**Leave a trail your AI agents can follow.**
+
+Persistent, local-first memory for coding agents over MCP. Tell your agent something once,
+and every future session remembers it.
+
+[![CI](https://github.com/RohitDeshmukh-1/Mindtrail-MCP/actions/workflows/ci.yml/badge.svg)](https://github.com/RohitDeshmukh-1/Mindtrail-MCP/actions/workflows/ci.yml)
+[![Python 3.11+](https://img.shields.io/badge/python-3.11%2B-blue)](https://www.python.org/)
+[![MCP](https://img.shields.io/badge/MCP-compatible-8A2BE2)](https://modelcontextprotocol.io)
+[![License: MIT](https://img.shields.io/badge/license-MIT-green)](LICENSE)
+
+</div>
+
+---
+
+Every new agent session starts from zero. You explain your conventions again, re-state your
+preferences, and re-describe decisions you made last week. **Mindtrail gives your agents a shared,
+long-term memory** through the [Model Context Protocol](https://modelcontextprotocol.io), so
+what one session learns, every later session (in any tool) can recall.
+
+- 🔌 **Works with any MCP client.** One memory shared by Claude Code, Cursor, VS Code, Codex
+  and your own agents.
+- 🏠 **Local-first and private.** One SQLite file on your machine. No account, no API key, no
+  telemetry.
+- ⚡ **Installs in seconds.** No model download is required; add neural embeddings later with
+  one extra.
+- 🗂️ **Scoped automatically.** Repo facts stay with the repo (detected from the git remote) and
+  personal preferences follow you everywhere.
+- 🔎 **Hybrid retrieval.** Keyword (BM25) and vector search, fused and ranked, with the ranking
+  signals shown for every result.
+- 🛡️ **Safe by default.** Refuses to store credentials, marks recalled memory as untrusted data,
+  and deletes for real.
+
+## Does it actually help?
+
+We tested it with real Claude Code sessions. In one session the user mentions a project fact
+in passing ("FYI, this project uses pnpm"). A fresh session later gets a task that depends on
+it ("how do I add lodash?"). The agent is never told to use Mindtrail.
+
+| | stored the fact on its own | applied it in a later session |
+|---|---:|---:|
+| Claude Code without Mindtrail | – | 0/12 |
+| Claude Code with `mindtrail[semantic]` | **36/36** | **36/36** |
+
+Without memory, the agent answers from habit every time: `npm install lodash`, a commit
+message in the wrong format, port 5432 instead of your 5433. With Mindtrail it checks first
+and gets your project's answer. This is a small test (12 facts run three times, one model,
+an empty repository), not a measure of task success on large codebases.
+[Method, raw results and how to reproduce](benchmarks/README.md#agent-evaluation).
+
+## Quickstart
+
+```bash
+pipx install "mindtrail[semantic]"                         # or: uv tool install "mindtrail[semantic]"
+mindtrail init                                             # downloads the embedding model once (~210 MB)
+claude mcp add mindtrail --scope user -- mindtrail serve   # Claude Code
+```
+
+Have [uv](https://docs.astral.sh/uv/)? Skip the install step; `uvx` fetches Mindtrail on first
+run (the embedding model downloads in the background the first time the server starts):
+
+```bash
+claude mcp add mindtrail --scope user -- uvx --from "mindtrail[semantic]" mindtrail serve
+```
+
+Cursor, VS Code, Codex and custom clients are covered in
+**[docs/integrations.md](docs/integrations.md)**, and `mindtrail init` prints each config.
+Want the smallest install? `pipx install mindtrail` skips the model download and matches on
+words instead of meaning (see [Better semantic recall](#better-semantic-recall)).
+
+Then try it:
+
+```text
+Session 1 › FYI, this project uses conventional commits and squash merges.
+Session 2 › Write a commit message for these changes.
+            → the agent recalls the convention and writes "feat(api): add pagination to /orders"
+```
+
+## Using it day to day
+
+You don't need special commands. Work as usual and the agent decides what to keep:
+
+- **Mention things once.** "We deploy from the `release` branch", "I prefer pytest over
+  unittest", "the flaky test was a timezone bug, fixed by pinning TZ=UTC". The agent stores
+  facts like these on its own. Say "remember that…" when you want to be sure.
+- **Project vs. personal.** Facts about the repository go to a project space, shared by every
+  clone of the same git remote. Facts about you ("I like short answers") go to your personal
+  space and apply in every project.
+- **Things change.** Say "we moved from npm to pnpm" and the agent replaces the old memory
+  instead of keeping both. The old one is kept as history but no longer recalled.
+- **Ask what it knows.** "What do you remember about this project?" or, from the terminal,
+  `mindtrail list` and `mindtrail recall "<question>"`.
+- **Forget anything.** "Forget the staging URL" in chat, or `mindtrail forget <id>`. Deletes
+  are permanent.
+- **Switch tools freely.** Claude Code, Cursor and Codex pointed at the same Mindtrail share
+  one memory, so a convention taught in one tool is known in all of them.
+
+Good things to store: conventions, commands, ownership, where config lives, decisions and why
+they were made, root causes of tricky bugs. Don't bother with what the code or git history
+already says. Secrets are refused automatically.
+
+## How it works
+
+Your agent gets three tools, and Mindtrail's server instructions tell it when to use them:
+
+| Tool | What it does |
+|---|---|
+| `remember` | Store one fact, preference, decision or event, scoped to the **project** or **personal** space. Pass `replaces` to supersede an outdated memory. |
+| `recall` | Find relevant memories from the current project plus your personal space. Returns nothing when nothing relevant exists. |
+| `forget` | Permanently delete a memory. |
+
+Set `MINDTRAIL_TOOLS=full` for four more: `search_memory` (filters by space, type and validity),
+`get_context` (a prompt-ready block within a token budget), `update_memory` (edits with
+version history) and `get_memory`.
+
+Behind the tools is a small, well-tested engine: duplicate merging, supersession, validity
+windows, version history and hybrid ranking. See **[docs/architecture.md](docs/architecture.md)**.
+
+## Manage memory from the terminal
+
+```bash
+mindtrail recall "how do we deploy?"   # search project + personal memory
+mindtrail remember "Staging is at staging.example.com" --scope project
+mindtrail list                         # newest first
+mindtrail forget <id>                  # permanent delete
+mindtrail export -o memories.jsonl     # everything you've stored, as JSON Lines
+mindtrail doctor                       # diagnose the install
+```
+
+## Better semantic recall
+
+The default embedder matches on words and word fragments. For recall by meaning ("how do we
+deploy?" → "deploys go through GitHub Actions"), install the local neural model. It runs on CPU
+and needs no API key:
+
+```bash
+pipx install "mindtrail[semantic]"
+mindtrail init        # downloads the model once (~210 MB) and re-indexes existing memories
+```
+
+For a smaller download (67 MB, somewhat lower recall), set
+`MINDTRAIL_EMBEDDING_MODEL=BAAI/bge-small-en-v1.5`. Memories stored under another model are
+re-embedded automatically the next time the server starts.
+
+## Retrieval benchmark
+
+The agent test above is in [Does it actually help?](#does-it-actually-help). On two held-out sets of developer-memory questions (82 in total) that were never used for
+tuning ([methodology](benchmarks/README.md)):
+
+| | recall@1 | recall@5 | paraphrase recall@5 | correctly says "nothing stored" | stale/foreign leaks |
+|---|---:|---:|---:|---:|---:|
+| default install | 58–64% | 68–75% | 44–46% | 82–100% | **0%** |
+| `mindtrail[semantic]` | **82–90%** | **93–94%** | **85–89%** | **100%** | **0%** |
+
+This is a small, synthetic retrieval benchmark written by us. Run `mindtrail bench` to
+reproduce it, or add your own dataset.
+
+## Use it from Python
+
+```python
+from mindtrail import MemoryService
+
+memory = MemoryService.from_config()
+memory.remember("The API uses FastAPI and PostgreSQL", space_id="project:shop")
+print(memory.get_context("add a new endpoint", space_ids=["project:shop"]).text)
+```
+
+## Privacy and security
+
+Everything stays in `~/.mindtrail/mindtrail.db` on your machine. Mindtrail refuses writes that
+look like API keys, tokens or private keys. `forget` overwrites deleted data on disk, and recalled memories
+are marked as untrusted reference data so agents don't follow instructions stored inside them.
+See [SECURITY.md](SECURITY.md) to report issues.
+
+## Roadmap
+
+- [x] Core engine: hybrid retrieval, dedup, supersession, validity windows, history
+- [x] MCP server and CLI, with automatic project scoping
+- [ ] PostgreSQL + pgvector backend with multi-tenant isolation
+- [x] Retrieval benchmark suite in CI (recall@k, MRR, abstention, leak rate)
+- [ ] Hosted remote MCP with one-click OAuth connectors
+- [ ] Web memory viewer (browse, edit, delete, export)
+- [ ] Entity graph, contradiction detection and memory consolidation
+
+## Contributing
+
+Issues and PRs are welcome. Read [CONTRIBUTING.md](CONTRIBUTING.md) to get set up; the whole
+suite runs in about 15 seconds. If Mindtrail saves you from re-explaining your codebase, a ⭐
+helps others find it.
+
+## License
+
+[MIT](LICENSE)
+
+<!-- mcp-name: io.github.rohitdeshmukh-1/mindtrail -->

mindtrail-0.1.0/README.md

@@ -0,0 +1,198 @@
+<div align="center">
+
+# 🧭 Mindtrail
+
+**Leave a trail your AI agents can follow.**
+
+Persistent, local-first memory for coding agents over MCP. Tell your agent something once,
+and every future session remembers it.
+
+[![CI](https://github.com/RohitDeshmukh-1/Mindtrail-MCP/actions/workflows/ci.yml/badge.svg)](https://github.com/RohitDeshmukh-1/Mindtrail-MCP/actions/workflows/ci.yml)
+[![Python 3.11+](https://img.shields.io/badge/python-3.11%2B-blue)](https://www.python.org/)
+[![MCP](https://img.shields.io/badge/MCP-compatible-8A2BE2)](https://modelcontextprotocol.io)
+[![License: MIT](https://img.shields.io/badge/license-MIT-green)](LICENSE)
+
+</div>
+
+---
+
+Every new agent session starts from zero. You explain your conventions again, re-state your
+preferences, and re-describe decisions you made last week. **Mindtrail gives your agents a shared,
+long-term memory** through the [Model Context Protocol](https://modelcontextprotocol.io), so
+what one session learns, every later session (in any tool) can recall.
+
+- 🔌 **Works with any MCP client.** One memory shared by Claude Code, Cursor, VS Code, Codex
+  and your own agents.
+- 🏠 **Local-first and private.** One SQLite file on your machine. No account, no API key, no
+  telemetry.
+- ⚡ **Installs in seconds.** No model download is required; add neural embeddings later with
+  one extra.
+- 🗂️ **Scoped automatically.** Repo facts stay with the repo (detected from the git remote) and
+  personal preferences follow you everywhere.
+- 🔎 **Hybrid retrieval.** Keyword (BM25) and vector search, fused and ranked, with the ranking
+  signals shown for every result.
+- 🛡️ **Safe by default.** Refuses to store credentials, marks recalled memory as untrusted data,
+  and deletes for real.
+
+## Does it actually help?
+
+We tested it with real Claude Code sessions. In one session the user mentions a project fact
+in passing ("FYI, this project uses pnpm"). A fresh session later gets a task that depends on
+it ("how do I add lodash?"). The agent is never told to use Mindtrail.
+
+| | stored the fact on its own | applied it in a later session |
+|---|---:|---:|
+| Claude Code without Mindtrail | – | 0/12 |
+| Claude Code with `mindtrail[semantic]` | **36/36** | **36/36** |
+
+Without memory, the agent answers from habit every time: `npm install lodash`, a commit
+message in the wrong format, port 5432 instead of your 5433. With Mindtrail it checks first
+and gets your project's answer. This is a small test (12 facts run three times, one model,
+an empty repository), not a measure of task success on large codebases.
+[Method, raw results and how to reproduce](benchmarks/README.md#agent-evaluation).
+
+## Quickstart
+
+```bash
+pipx install "mindtrail[semantic]"                         # or: uv tool install "mindtrail[semantic]"
+mindtrail init                                             # downloads the embedding model once (~210 MB)
+claude mcp add mindtrail --scope user -- mindtrail serve   # Claude Code
+```
+
+Have [uv](https://docs.astral.sh/uv/)? Skip the install step; `uvx` fetches Mindtrail on first
+run (the embedding model downloads in the background the first time the server starts):
+
+```bash
+claude mcp add mindtrail --scope user -- uvx --from "mindtrail[semantic]" mindtrail serve
+```
+
+Cursor, VS Code, Codex and custom clients are covered in
+**[docs/integrations.md](docs/integrations.md)**, and `mindtrail init` prints each config.
+Want the smallest install? `pipx install mindtrail` skips the model download and matches on
+words instead of meaning (see [Better semantic recall](#better-semantic-recall)).
+
+Then try it:
+
+```text
+Session 1 › FYI, this project uses conventional commits and squash merges.
+Session 2 › Write a commit message for these changes.
+            → the agent recalls the convention and writes "feat(api): add pagination to /orders"
+```
+
+## Using it day to day
+
+You don't need special commands. Work as usual and the agent decides what to keep:
+
+- **Mention things once.** "We deploy from the `release` branch", "I prefer pytest over
+  unittest", "the flaky test was a timezone bug, fixed by pinning TZ=UTC". The agent stores
+  facts like these on its own. Say "remember that…" when you want to be sure.
+- **Project vs. personal.** Facts about the repository go to a project space, shared by every
+  clone of the same git remote. Facts about you ("I like short answers") go to your personal
+  space and apply in every project.
+- **Things change.** Say "we moved from npm to pnpm" and the agent replaces the old memory
+  instead of keeping both. The old one is kept as history but no longer recalled.
+- **Ask what it knows.** "What do you remember about this project?" or, from the terminal,
+  `mindtrail list` and `mindtrail recall "<question>"`.
+- **Forget anything.** "Forget the staging URL" in chat, or `mindtrail forget <id>`. Deletes
+  are permanent.
+- **Switch tools freely.** Claude Code, Cursor and Codex pointed at the same Mindtrail share
+  one memory, so a convention taught in one tool is known in all of them.
+
+Good things to store: conventions, commands, ownership, where config lives, decisions and why
+they were made, root causes of tricky bugs. Don't bother with what the code or git history
+already says. Secrets are refused automatically.
+
+## How it works
+
+Your agent gets three tools, and Mindtrail's server instructions tell it when to use them:
+
+| Tool | What it does |
+|---|---|
+| `remember` | Store one fact, preference, decision or event, scoped to the **project** or **personal** space. Pass `replaces` to supersede an outdated memory. |
+| `recall` | Find relevant memories from the current project plus your personal space. Returns nothing when nothing relevant exists. |
+| `forget` | Permanently delete a memory. |
+
+Set `MINDTRAIL_TOOLS=full` for four more: `search_memory` (filters by space, type and validity),
+`get_context` (a prompt-ready block within a token budget), `update_memory` (edits with
+version history) and `get_memory`.
+
+Behind the tools is a small, well-tested engine: duplicate merging, supersession, validity
+windows, version history and hybrid ranking. See **[docs/architecture.md](docs/architecture.md)**.
+
+## Manage memory from the terminal
+
+```bash
+mindtrail recall "how do we deploy?"   # search project + personal memory
+mindtrail remember "Staging is at staging.example.com" --scope project
+mindtrail list                         # newest first
+mindtrail forget <id>                  # permanent delete
+mindtrail export -o memories.jsonl     # everything you've stored, as JSON Lines
+mindtrail doctor                       # diagnose the install
+```
+
+## Better semantic recall
+
+The default embedder matches on words and word fragments. For recall by meaning ("how do we
+deploy?" → "deploys go through GitHub Actions"), install the local neural model. It runs on CPU
+and needs no API key:
+
+```bash
+pipx install "mindtrail[semantic]"
+mindtrail init        # downloads the model once (~210 MB) and re-indexes existing memories
+```
+
+For a smaller download (67 MB, somewhat lower recall), set
+`MINDTRAIL_EMBEDDING_MODEL=BAAI/bge-small-en-v1.5`. Memories stored under another model are
+re-embedded automatically the next time the server starts.
+
+## Retrieval benchmark
+
+The agent test above is in [Does it actually help?](#does-it-actually-help). On two held-out sets of developer-memory questions (82 in total) that were never used for
+tuning ([methodology](benchmarks/README.md)):
+
+| | recall@1 | recall@5 | paraphrase recall@5 | correctly says "nothing stored" | stale/foreign leaks |
+|---|---:|---:|---:|---:|---:|
+| default install | 58–64% | 68–75% | 44–46% | 82–100% | **0%** |
+| `mindtrail[semantic]` | **82–90%** | **93–94%** | **85–89%** | **100%** | **0%** |
+
+This is a small, synthetic retrieval benchmark written by us. Run `mindtrail bench` to
+reproduce it, or add your own dataset.
+
+## Use it from Python
+
+```python
+from mindtrail import MemoryService
+
+memory = MemoryService.from_config()
+memory.remember("The API uses FastAPI and PostgreSQL", space_id="project:shop")
+print(memory.get_context("add a new endpoint", space_ids=["project:shop"]).text)
+```
+
+## Privacy and security
+
+Everything stays in `~/.mindtrail/mindtrail.db` on your machine. Mindtrail refuses writes that
+look like API keys, tokens or private keys. `forget` overwrites deleted data on disk, and recalled memories
+are marked as untrusted reference data so agents don't follow instructions stored inside them.
+See [SECURITY.md](SECURITY.md) to report issues.
+
+## Roadmap
+
+- [x] Core engine: hybrid retrieval, dedup, supersession, validity windows, history
+- [x] MCP server and CLI, with automatic project scoping
+- [ ] PostgreSQL + pgvector backend with multi-tenant isolation
+- [x] Retrieval benchmark suite in CI (recall@k, MRR, abstention, leak rate)
+- [ ] Hosted remote MCP with one-click OAuth connectors
+- [ ] Web memory viewer (browse, edit, delete, export)
+- [ ] Entity graph, contradiction detection and memory consolidation
+
+## Contributing
+
+Issues and PRs are welcome. Read [CONTRIBUTING.md](CONTRIBUTING.md) to get set up; the whole
+suite runs in about 15 seconds. If Mindtrail saves you from re-explaining your codebase, a ⭐
+helps others find it.
+
+## License
+
+[MIT](LICENSE)
+
+<!-- mcp-name: io.github.rohitdeshmukh-1/mindtrail -->