Doriku
Doriku is a shared control plane for AI coding agents — Claude Code, Cursor, Windsurf, and human teammates sync tasks, memory, decisions, and file locks…
What it can do
- Doriku Task: Tasks: lifecycle, coordination, recovery. create with objective+done_criteria and log with progress_summary+next_step are quality-scored; resume returns a full resume pack.
- Doriku Agent: Agent presence and coordination. heartbeat responses include pending_commands[] (console-issued interrupt/inject_task) — act on them when present.
- Doriku Memory: Persistent memory shared across sessions, projects, and AI tools. relate/neighbors manage the memory graph; glossary_*/preference_* tune search scoring.
What data it sees
Do you need an account
No: the server works without sign-in
Doriku is a shared control plane for AI coding agents — Claude Code, Cursor, Windsurf, and human teammates sync tasks, memory, decisions, and file locks through one MCP server.\n\n12 consolidated tools (~15KB tools/list) covering:\n\n- Tasks & coordination — atomic claim/release, dependencies, resume packs that restore full context across sessions and tools\n- Persistent memory — workspace-scoped knowledge with lexical + semantic search\n- Governance — file lock policies, cost caps, approval workflows, audit trail\n- Automations & workflows — schedules, status triggers, Slack/Discord notifications\n- Team visibility — live console dashboard, agent presence, handoff summaries\n\nMulti-agent safe by design: two agents can work the same repo without stepping on each other.\n\nGet started: sign up at https://doriku.io (free tier: 2 agents, 100 tasks/mo), then npx @doriku/cli setup. Auth: Bearer API key from the console.
Server tool list (12)
Raw names from tools/list. Only developers need these.
| doriku_task | Tasks: lifecycle, coordination, recovery. create with objective+done_criteria and log with progress_summary+next_step are quality-scored; resume returns a full resume pack. |
| doriku_agent | Agent presence and coordination. heartbeat responses include pending_commands[] (console-issued interrupt/inject_task) — act on them when present. |
| doriku_memory | Persistent memory shared across sessions, projects, and AI tools. relate/neighbors manage the memory graph; glossary_*/preference_* tune search scoring. |
| doriku_context | Key-value working context (workspace/project/agent scoped). set with source+scope records provenance. |
| doriku_decision | Decision records. record with problem+chosen_option+alternatives+rationale is quality-scored. |
| doriku_project | Projects group related tasks. summarize generates a progress summary; summary returns the latest. |
| doriku_workspace | Workspaces and state sync: snapshot at session start, then poll changes with the returned sequence number. |
| doriku_workflow | Multi-step workflows across agents. step_* verbs manage step run lifecycle for executing agents. |
| doriku_automation | Automation rules (triggers: task_status_change|task_created|schedule|manual) and the Smart Dispatcher (Pro). |
| doriku_governance | Multi-agent guardrails: file locks, lock policies, token cost caps, audit trail. |
| doriku_search | Global search across tasks, projects, and agents. Top 5 per category by relevance. |
| doriku_session | Session flow: structured handoffs, next actions, token usage reporting, plan capabilities. |