Memory & Context

Search past AI conversation history with the trellis mem CLI for prior decisions

Mine past transcripts and update AGENTS.md so agents learn from earlier work

Turn preferences from past chats into reusable skills, rules and workflow docs

Answer questions from local knowledge base folders with PDF and Excel retrieval

Show what AI coding tools cost today by model, project and task

Build a persistent Markdown wiki from sources that you can query and lint

Give your assistant a two-tier memory that decodes shorthand, nicknames and acronyms

Save critical values from tool results before context pruning drops them

Keep a running scratchpad of decisions, values and open questions across a task

Build, query and inspect a project knowledge graph stored in the planning folder

File phase artifacts into MemPalace and mirror decisions into its knowledge graph

Recall past decisions, patterns and surprises from MemPalace before planning

Write a context handoff when you pause work partway through a phase

Build a developer behavior profile that agents can read to adapt to you

Resume a previous session with its full context restored

Manage persistent context threads for work that spans many sessions

Turn a local folder into a searchable knowledge base agents can keep up to date

Restore the last saved diagnosis state to pick up where you left off

Save key diagnosis state locally and manage where saved files are stored

Save and distill lessons from each session into reusable notes

Keep the newer half of a long conversation to save tokens and keep recent work

Write a handoff document so the next session can pick up the work cleanly

Keep only the last quarter of a long conversation to cut token use

Write a session log of completed work, findings and open questions

Search earlier session transcripts to recover past findings and resume work

Read shared memory and inbox messages at task start and record durable facts

Search raw past chat messages by keyword, channel and time when memory is not enough

Operate agent memory with pending publication, merged search and snapshots

Restore the working picture of a coding agent after context compaction or a reset

Design how agent context windows get filled, pruned and shared

Load one section of a Markdown file by heading anchor for agent context

Build a world model after taking over an agent seat through a planned handover

Re-anchor a long-running agent on its product goal after compaction or a boundary

Design restore packets so an agent session survives compaction and restarts

Start a new agent seat by forking an earlier coding agent session

Inspect Honcho workspaces, peer representations and memory with the honcho CLI

Set Honcho peer and session naming so several Grok bots share one user memory

Give agents persistent memory across conversations with Honcho saved user context

Set up and health-check Honcho memory in Cursor or Grok Bot, including sign-in

Check and use provided context files before searching externally

Read provided reference files before searching the web or guessing data

Always use provided reference data before any outside search

Review auto-memory entries and propose promotions while catching duplicates and conflicts

Extract data from context files with validation before making assumptions

Answer quick side questions from the current session without tool calls

Write a handoff note so the next session can resume exactly where work stopped

Diagnose bloated or brittle AI workflows and fix context stuffing problems

Store customer traits and past conversations in Twilio Memory for agents and staff

Save product sources and design preferences so later sessions start with context