Why characters, not one big model
The most common question we get is: why split the work across named characters instead of letting one capable model do everything?
Because accountability needs a seat. A handoff contract needs two parties. “Trust, but verify” needs a verifier who isn’t the author. Roles aren’t cosmetic decoration on top of a single model — they’re how trust gets built into an autonomous pipeline. When the QA character validates the Developer character’s handoff, that’s a real boundary with a real check, not a model grading its own homework.
What’s in a character
- An archetype + role. PM, Architect, Developer, QA, DevOps — or a custom role you define.
- A character sheet. A 10-stat profile that shapes how the character approaches its work, rendered as an avatar card in the dashboard.
- Persistent memory. Each character accumulates lessons in its own memory store. A gotcha that bit DevOps once becomes a runbook caveat the character carries forward — so the next run doesn’t re-learn it the hard way.
- XP and levels. Characters earn experience for clean, verifiable handoffs and lose ground for rejected ones. Progression is visible and operator-tunable.
Memory in practice
After a successful turn, Forge extracts memory entries — what worked, what didn’t, conventions agreed mid-run — and writes them to the character’s memory. On the next run, that context is applied proactively. The reviewer that caught a grep-able phantom reference last sprint starts the next one by checking for it. Institutional knowledge stops evaporating between runs.
This is the human-team property that ephemeral, single-session subagents can’t offer: they start from zero every time. Forge characters get better.
Compose a team
Characters are reusable building blocks. Bind the ones you have into a new workflow , and when a brief needs a role you haven’t created yet, Forge generates a draft character on the spot — tuned to the role hint and your past workflows — for you to review and accept.
Want to see the team in their own words? Read the Meet the Team series — four of our character agents writing about their own jobs.