MCP SERVERStore the way you work as a step-by-step workflow, so a new session picks it up instead of asking you to explain the process again.
LABSCO SUMMARY
The work loop is deliberately small: ask for the current step, get its playbook and output spec, submit the output, get the next step. Because workflows and task state live in a database rather than in the conversation, a context reset costs you nothing, and a playbook you sharpen after a bad run applies to every task afterwards.
A workflow engine that keeps your processes in a local SQLite file. Each workflow is an ordered set of steps, each step carries a playbook and an output spec, and the agent pulls one step at a time.
No account and no key — everything lives in a SQLite file on your machine. Run it with uvx progi from the PyPI package progi over stdio; the published version is 0.3.2. The bundled mode starts the MCP server and the monitoring UI together; uvx progi --no-web or PROGI_NO_WEB=1 runs the server alone, and the docs run the UI alone under the name progi-web, which PyPI does not have. PROGI_DB_PATH moves the database file — use an absolute path — while PROGI_WEB_HOST and PROGI_WEB_PORT move the UI off 127.0.0.1:8000.
One command — uvx progi
USE CASES
Start the task, work the steps in order against your own playbooks, and submit each step's output before the next one opens.
Use the skeleton prompt to break the description into steps, write a playbook per step, then persist both.
Ask for the task again — it returns the step you were on, its input data and your saved progress notes.
Replace that step's playbook; every task started afterwards uses the new one.
Save progress notes on the task; they come back with the step context and clear themselves when the step completes.
Counted once with only startup setup. Other versions, modes or optional credentials may expose a different list.
create_taskCreate a new task under the given workflow.
Creates the task (status 'todo') and returns the task plus a preview of its first step. Before calling this, confirm the workflow choice with the user AND ask them what they want to name the task.
get_playbook_authoring_promptReturn the Pass 2 system prompt for authoring a step's playbook.
Workflow context (the full process, this step's position, and its input/output specs) is injected at the top of the prompt template.
get_process_skeleton_promptReturn the Pass 1 system prompt for authoring a new workflow's skeleton.
The harness uses it to help the user convert a plain-language workflow description into a structured process skeleton (steps with input/output specs, no playbooks yet).
list_tasksList tasks, optionally filtered by status and/or workflow_id.
Empty string / 0 means no filter. "My todos" = status="todo".
list_workflowsReturn all workflows with their ordered steps.
save_workflowPersist a new workflow, its steps, and playbooks.
skeleton: the JSON object produced by Pass 1 (process skeleton prompt). playbooks_by_step: mapping of step name → playbook markdown string.
start_or_continue_taskMain work-loop entry point.
Context includes task info, the current step name + position, input_data, output_spec (the expected format/type of the deliverable), the playbook markdown, and progress_notes (if any).
Before calling submit_output, verify that your output satisfies output_spec (correct type, meets constraints, includes any fields referenced by branching conditions).
submit_outputMark the current step complete, store its output, and advance.
Either returns the next step's info (name + playbook, so the agent can continue immediately) or {'status': 'done'} if it was the last step.
Pass task_name to rename the task when you now have enough context to give it a meaningful name (e.g. after the first step reveals what the task is actually about). Leave it empty to keep the current name.
IMPORTANT — approval gate: if start_or_continue_task returned current_step.requires_approval = true for this step, you MUST present the output to the user and ask for explicit approval BEFORE calling this tool. Only call submit_output once the user has confirmed they are happy with the output. If they request changes, make them first, then ask again.
update_playbookReplace the playbook content for a step.
update_progress_notesOverwrite a task's progress_notes.
Only call this when the user explicitly asks to save or update progress notes. Notes are cleared automatically when a step completes.
BEFORE YOU INSTALL
10 tools when we connected. The documented finish_step is not exposed and appears to have become submit_output. update_playbook, also undocumented, is the only in-agent way to tweak playbooks between runs.
uvx on PATH
CONTEXT COST
Rated by the size of the 10 tool definitions that ride along on every turn, not by how many there are
INSTALLATION
Add the block to your client, or run the command directly.
{
"mcpServers": {
"progi": {
"command": "uvx",
"args": [
"progi"
]
}
}
}uvx progiMCP servers
EchoingVesper
BAM-DevCrew
flytohub