DogearAI Memory
DogearAI is the user's own AI memory layer: one place that their Claude, Cursor, ChatGPT and other tools all share. This skill lets you recall what the user has already saved before you work, and save durable new facts so they never have to re-explain themselves in a different tool.
Everything runs through the bundled dogear.py script — zero setup, no signup. The
first call auto-creates a DogearAI account and saves its token locally; nothing to learn,
no MCP server to configure.
Setup — none
Just call it. On first use the script auto-creates an anonymous DogearAI account and saves
its token to ~/.dogear/token. To keep the memories (add an email, use them on other
machines), open the claim link printed on first run, or run python dogear.py login to
see it.
Power users can bring their own token instead: python dogear.py set-token dg_xxx, or set
DOGEAR_TOKEN=dg_xxx (takes precedence).
How to call it
Run the bundled dogear.py (stdlib Python, no install) with Bash. Use its path inside
this skill's folder.
| Goal | Command |
|---|---|
| Pull the user's context | python dogear.py context (opt. --scopes a,b, --max-tokens N) |
| Save a memory | python dogear.py remember "<the user's words>" --source chat |
| List memory spaces | python dogear.py spaces |
| Read one space in full | python dogear.py read-space <space_id> |
| Fetch a memory's raw original | python dogear.py get <memory_id> |
| Save an API token | python dogear.py set-token dg_xxx |
| Sign-in / token help | python dogear.py login |
When to recall (pull context first)
At the start of a task where the user's long-term context would help — they mention
"my project", "my preferences", "the stack", "as we decided", or you're resuming earlier
work — run python dogear.py context before asking them to repeat anything.
- Default pulls the navigation index + all
activespaces. - Know the area? Add
--scopes a,b, or usespaces→read-spaceto read one in full. - Recalled memory reflects what was true when written. Verify anything that may have changed (a file, a version, a decision) before relying on it.
When to save (write a memory)
Whenever the user states something durable they'd hate to re-explain in another tool,
run python dogear.py remember "<their words>".
Save: preferences ("I prefer X", "always Y"), decisions ("we're going with X", "we ruled out Y"), facts about the user / project / stack / goals, commitments and TODOs.
Don't save: transient chatter, anything trivially re-derivable from the current files/code, or secrets — unless the user asks.
The server classifies the memory and files it into the right space for you — you don't pick the space. One call = one atomic memory; split long, multi-topic notes into several calls. A short "Saved to DogearAI" is enough.
Etiquette
- Reads (
context,spaces,read-space) are safe and silent — no need to ask. - Every call is scoped to the user's own account by their token — you never pass a user id.
- Be selective when saving: a few high-value memories beat dumping everything.
- Curation (renaming, merging, deleting spaces) is done by the user in the DogearAI dashboard, not through this skill.
Already use the DogearAI MCP server? Its tools (get_context / list_spaces / read_space / write_memory) do the same thing — this script is just the CLI path.