
Going Anyway? Building ErrandLoop with AWS’s Open-Source Cedar
How a neighbourhood pickup prototype matches existing trips into reciprocal exchanges, with AWS’s open-source Cedar enforcing participant permissions. Lessons from building the matching workflow, clarifying handovers and testing a three-person demonstration.
Going Anyway? Building ErrandLoop with AWS’s Open-Source Cedar
A small favour can involve surprisingly complicated coordination.
Someone needs a library pickup. Another person is heading to the grocery shop. A third is collecting printouts. Their trips already exist, but figuring out who can help whom requires comparing destinations, timing and carrying space.
For First Commit, I built ErrandLoop to explore whether software could make those exchanges easier to understand and complete.
Why a circle?
A direct swap only works when two people can help each other.
Consider three neighbours:
- Asha is going to the print shop and needs a library pickup.
- Ravi is going to the library and needs groceries.
- Meena is going to the grocery shop and needs a printout.
No pair can satisfy both people’s requests. A three-person circle can: Asha collects for Meena, Meena for Ravi, and Ravi for Asha.
ErrandLoop finds reciprocal circles of two to four participants. It checks shared pickup locations, readiness, return deadlines and carrying capacity. Among compatible possibilities, it selects non-overlapping circles that serve the most members.
It does not calculate road routes or verify traffic conditions. Its decisions use the plans participants enter.
Where AWS Cedar fits
An exchange is only useful if people can trust its permissions.
Who may accept a proposal? Who may mark an item collected? Who may confirm receipt?
ErrandLoop uses AWS’s open-source Cedar authorization engine through the separately maintained cedarpy Python binding. The backend evaluates participant permissions before protected actions. Application logic separately enforces the exchange’s state transitions.
That distinction matters: being an authorized participant does not mean an action is valid at every stage.
Everyone must accept before collection begins. Only the relevant collector can record a pickup, and only its recipient can confirm receipt.
The runnable local application uses Python, SQLite and a JavaScript interface. This submission does not claim a live AWS deployment.
Designing around responsibility
My design focus became a practical question: can someone quickly understand what they need to do?
The interface now shows a structured exchange panel and personal “You collect” and “You receive” summaries. A progress indicator follows Agree, Collect and Hand over.
Earlier versions used a compact triangle diagram. Reviewing screenshots exposed crowded labels and unclear relationships. I replaced it with numbered handover rows containing the collector, recipient, location and item.
Other refinements addressed oversized cards, malformed avatar initials, unclear empty states and form errors. Invalid times now produce an immediate error without discarding entered notes.
Handling plans that change
Cancellation is more complicated once an item has been collected.
Before collection, available posts can be released. After collection, ErrandLoop preserves the outstanding handover and avoids automatically matching the item again.
The prototype does not provide substitute collectors or dispute resolution. It keeps the unfinished responsibility visible.
Testing without overstating the results
Validation includes 57 passing Python tests and automated HTTP/DOM checks. These exercise account creation, invitations, consent, collection, recipient confirmation and cancellation recovery.
For the demonstration, an isolated rehearsal mode prepares three test accounts and their posts. A localhost-only switcher lets me demonstrate all participants in one window. Normal accounts retain their ordinary sign-in flow.
These are simulated checks, not independent usability research. In-app confirmations are participant reports, and suggested trip savings are estimates—not measured environmental or financial impact.
What I learned
The most useful interface improvements came from making responsibility explicit. A prominent button is not enough if a person cannot tell what they are agreeing to.
ErrandLoop is a working prototype for small, trusted groups. The next meaningful validation is observing people use it for actual planned pickups and learning where coordination still breaks down.
Disclosure: ChatGPT/Codex assisted with design, implementation, testing and writing. The demonstration uses simulated pickups.
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article