data-ai MCP Server
Local-first AI agent bootstrap: Playwright Browser MCP + ContextDB for Codex CLI, Claude Code, Gemini CLI, and OpenCode.
Discovered via agent-topic:ai-agent and last synced 3mo ago.
Install instructions not detected yet
Check the source repository for the latest setup steps.
gemini-cli>`) - Note: in-CLI reset commands like `/new` (Codex) or `/clear` (Claude/Gemini) reset the conversation state. Exit and re-run the CLI to re-inject, or attach `memory/context-db/exports/latest-<agent>-context.md` as the first prompt. - Limitation: does not auto-write checkpoint after every turn ### B) One-shot mode (full automation recommended) ```bash scripts/ctx-agent.sh --agent codex-cli --project RexCLI --prompt "Continue from previous task and execute next step" ``` In one-shot mode, all 5 steps run automatically: `init -> session:new/latest -> event:add -> checkpoint -> context:pack` One-shot checkpoints also refresh compact continuity artifacts at `memory/context-db/sessions/<session_id>/continuity-summary.md` and `memory/context-db/sessions/<session_id>/continuity.json`. Future context packets and lazy startup prompts include this continuity summary so a new/cleared CLI can resume without rereading the full event history first. ## ContextDB Layout (L0/L1/L2) ```text memory/context-db/ manifest.json index/context.db index/sessions.jsonl index/events.jsonl index/checkpoints.jsonl sessions/<session_id>/ meta.json l0-summary.md continuity-summary.md continuity.json l1-checkpoints.jsonl l2-events.jsonl state.json exports/<session_id>-context.md ``` In global mode, this structure is created per project under that project's git root. ## Common Commands ```bash cd mcp-server npm run contextdb -- init npm run contextdb -- session:new --agent claude-code --project RexCLI --goal "stabilize flow" npm run contextdb -- event:add --session <id> --role user --text "need retry plan" npm run contextdb -- checkpoint --session <id> --summary "blocked by auth" --status blocked --next "wait-login
Description