Walkthrough

Step-by-step paths through owlcoda’s core workflows. Each path shows how the harness is entered, used, and carried forward.

Path 1 — Starting from zero

You have an idea but no structured project yet. This path shows how owlcoda creates initial project context from a rough description.

Step 1

Describe your intent

You tell owlcoda what you’re trying to build or solve. This can be as rough as a paragraph or as structured as a brief.

Task Definition
goalBuild the front-door HTTP service
typecoding
statusdraft
Step 2

Context is created

owlcoda generates a structured context: intent, constraints, open questions, and initial work items. This becomes the handoff document for all future sessions.

Context Created
specs3 accepted
scope6 files in scope
constraints2 recorded
statusready for dispatch
Step 3

Begin execution or hand off

You can start working immediately, or close the session knowing the context is preserved for later — whether you or someone else picks it up.

Dispatch Ready
taskimpl-front-door-mvp
assignedcopilot
worktreeisolated • clean branch
actiondispatch / edit / hold

Path 2 — Picking up an existing project

Someone else started this project, or you did but lost track. This path shows how owlcoda builds a handoff context from existing materials.

Step 1

Point to existing materials

Provide owlcoda with whatever exists: repos, docs, notes, previous chat exports. It doesn’t need everything — it works with what’s available.

Workspace Detection
repo/Users/demo/project
.owlcoda/found
tasks4 total • 2 pending review
Step 2

Structured handoff context

owlcoda produces a handoff summary: what’s been done, what’s pending, key decisions, and recommended next actions.

Handoff Context
completed2 tasks accepted
pending1 dispatched • 1 returned
decisions8 recorded
nextreview returned task
Step 3

Oriented and ready

You now have structured awareness of the project state. You can ask questions, start executing, or review specific areas.

Ready
statusoriented • review pending
actionsreview / dispatch / continue

Path 3 — Returning to continue work

You were here before. You took a break. Now you’re back. This is the path owlcoda is optimized for.

Step 1

Open your project

Open the project context. owlcoda immediately shows you the current state — no scrolling through old messages.

State on Return
since last session2 tasks completed
pending review1
new decisions3
Step 2

Review what changed

See what happened since your last session — completed items, new decisions, anything that shifted while you were away.

Changes Since Last Session
impl-executor-mvpreturned → accepted
spec-public-readmedispatched → returned
files changed12 across 2 tasks
Step 3

Continue from where you stopped

Pick up the next pending action. The context is already loaded — you’re productive in seconds, not minutes.

Resume
next actionreview spec-public-readme
contextloaded • ready

Path 4 — Review

You want to inspect the current state without changing anything. Review is the read-only path.

Step 1

Enter review mode

Request a review of the project. owlcoda presents the structured state: completed work, in-progress items, pending decisions, and blockers.

owlcoda review surface showing returned tasks with structured return document
Step 2

Drill into specifics

Inspect individual decisions, work items, or context areas. Everything is structured — no searching through chat logs.

Return Document Detail
summaryFront-door MVP implemented
files changed8
self-check4/5 passing
risks2 noted
actionaccept / reject / requeue

Path 5 — Execution

You’re ready to move work forward. Execution is the action path.

Step 1

Select a work item

Choose from pending actions. owlcoda shows what’s ready, what’s blocked, and what depends on what.

owlcoda execution view with dispatched task and context packet
Step 2

Execute with context

Work happens within the structured context. The model (or you) has full awareness of the project state, constraints, and history.

Execution
taskimpl-front-door-mvp
agentcopilot
worktreeisolated • active
statusin_progress
Step 3

Context updates automatically

When execution completes, the project context updates: items are marked done, new items may be created, and the handoff state reflects reality.

Task Returned
statusreturned → awaiting review
return docready
self-check4/5 passing
actionreview now / defer

Path 6 — Reset

A direction didn’t work, or a phase is complete. Reset lets you close out and start fresh without losing history.

Step 1

Initiate reset

Tell owlcoda you want to reset the current phase. It asks you to confirm and optionally note why.

Reset Confirmation
phasecurrent → archived
historypreserved
decisionscarried forward
Step 2

History preserved, slate cleared

The previous phase is archived — decisions and learnings are kept. But the active work context is clean, ready for a new direction.

After Reset
active tasks0
archived phases1
statusclean • ready for new work

Path 7 — Multi-model handoff

You’ve been working with one model and want to switch to another — or use multiple models for different parts of the work.

Step 1

Work with Model A

You complete a phase of work using one model. The project context captures everything that happened — not just the chat, but the structured state.

Session — Claude Code
taskimpl-front-door-mvp
agentclaude-code (coordinator)
actionsspec breakdown • 3 tasks created
Step 2

Hand off to Model B

Open a new session with a different model. Load the same project context. Model B has full awareness of what Model A did, without needing the original conversation.

Handoff — Copilot
taskimpl-front-door-mvp
agentcopilot (executor)
contextfull spec + scope • portable
previousclaude-code • 3 decisions carried
Step 3

Continuity without lock-in

You’re not dependent on any single model’s memory or conversation history. The continuity lives in the structured context, not in the model.

Continuity
models usedclaude-code, copilot
contextintact across handoff
lock-innone • model-agnostic

Want to know what’s working today and what’s coming next?

View Capabilities →