You decideAgents saveStateDockNext agent readsBuilds and records
Arrows show record handoffs, not automatic scheduling.
01 / CAPTURE
Plan in a conversation. Continue in your coding tools.
Changing devices or tools should not mean explaining the goal all over again.
Before you begin
Connect your conversation tool and coding agent to the same StateDock project with appropriate access. Mobile MCP access depends on your client and account; you can also save a plan through the StateDock mobile interface.
1Save the goal and decisions
2Ask your coding agent to read
3Build and record the outcome
The next agent can read the saved goal, scope, and outcome.
What to expectExternal tools and agents configure services, read screenshots, and write code. Shared records do not guarantee parameter accuracy or replace secret management.
03 / VISIBILITY
Multiple agents. One shared view.
With several tools at work, it is hard to see what is complete and what needs your decision.
Before you begin
Connect each agent to the same project and ask it to keep records up to date. Notifications require a separately configured supported channel.
1Share goals and tasks
2Agents record their progress
3Read outcomes and open questions
Find recorded progress, delivery notes, and blockers in one project.
What to expectThis view shows project state written back by agents, not live monitoring of external execution. Unrecorded progress does not appear automatically.
07 / REPORTING
Your work is already documented.
When several projects move at once, writing an update should not require reconstructing everything you delivered.
Before you begin
Tasks and completion summaries must be recorded in StateDock. Connect your AI through MCP with read access to the chosen projects. Define the date range, timezone, and project scope before reading the records.
1Choose dates and projects
2Read tasks, summaries, and history
3Ask AI to report by project
Use the same history for daily, weekly, and monthly reports—or combine the projects you specify.
What to expectYour connected AI generates reports on demand. Coverage depends on the records it can read. Scheduled reports and automatic delivery need external scheduling and sending tools.
Daily update
Completed work, current progress, and blockers
Weekly review
Key outcomes, project progress, unfinished work, and next priorities
Monthly summary
Milestones, product evolution, persistent backlog, and next direction
Multiple projects and agents, one review. Choose the project scope; your AI groups recorded outcomes by project.
Ask your connected AI
Use StateDock task records and completion summaries for the past 7 days to generate my development weekly review. Establish the exact dates, timezone, and project scope first; ask me if projects are unspecified. Read only the specified projects you have access to, and group a multi-project report by project.
Focus on: key outcomes, project progress, unfinished work, and next priorities. Include task references for actual outcomes and omit minor operations.
Read relevant tasks, summaries, and necessary history. Do not rely only on a short recent activity feed or a resume summary. Use completion timestamps or completion events for delivery dates, not last-edit timestamps. Label progress and blockers as current records rather than pretending they are snapshots at a historical cutoff.
State coverage limits if records are missing, dates cannot be verified, responses are truncated, or reads are incomplete. Do not invent outcomes, metrics, or unrecorded work. Label future priorities as suggestions and distinguish confirmed decisions. Generate the report only; do not schedule delivery.
GO FURTHER
One shared hub. More ways to work.
These approaches need additional agent configuration or work. They are not promises of automatic execution.
Use your connected agents together
Plan with one agent. Build with another.
Save scope and decisions with the tool you use for planning. Ask another connected agent to read and execute. StateDock shares the records; model choice, runtime, costs, and background execution belong to external tools.
Cloudflare Remote MCP · Bound identity required
Claim tasks with an agent identity.
Assign a task to a credential-bound identity and mark it READY. An agent can request its own queue. Identities do not create or wake agents. Claiming a task does not lock code files or prevent conflicts across tasks. Independent local adapters may behave differently.
Extension · Not a native one-click mirror
Create a translated project showcase.
An external agent with appropriate access can read records, translate them, and write a separate project. Translation, mapping, redaction, and ongoing synchronization require additional work. A showcase copy is not the original project audit history.
BEFORE YOU CONNECT
Your tools execute.
StateDock keeps the records.
A shared hub for tasks, project state, and handoff context.
Does StateDock run agents automatically?
No. StateDock stores shared state and context. Users and external hosts must initiate or separately configure execution, polling, or wakeups.
Does it synchronize whole conversations?
MCP does not listen to chats. Only plans, decisions, and outcomes explicitly written by you or a connected agent become project records.
Can I continue work from my phone?
You can save a plan using the StateDock mobile interface. MCP read/write access in a conversation client depends on your account, version, and authorization. A connected desktop agent can then read the same project.
Do progress and notifications appear automatically?
Agents need to write their updates. Notifications depend on supported channel configuration and delivery; immediate phone delivery is not guaranteed.
Can a different agent review the work?
You can ask another agent to read the handoff and record an independent review. A review record does not automatically trigger a review or mean that every completed task has been approved.
START WITH YOUR NEXT TASK
Give the next handoff a place to start.
Connect your tools. Keep goals, decisions, and outcomes in the same project.
Shared records. No automatic execution.