fig. 01—briefboard dispatch
Your AI agents, fully briefed.
Connect Claude Code over MCP: agents pull their next task, read the specs, ask in the technical channel, and report back — all on your project board.
↳ Dispatch is Briefboard's AI layer — included free in every plan
$ claude
› get_next_task
✓ BB-T204 — Add webhook retries (priority: high)
› get_milestone_specs
✓ 2 specs loaded — billing-webhooks v3
› update_task_status in_progress
● agent working — reporting to #technical
fig. 02—the dispatcher loop
Your agent asks for work. Briefboard answers.
No copy-pasting context into prompts. The agent pulls prioritized work, gets the full brief, and reports as it goes — you watch it happen on the board.
Pull
get_next_task returns the highest-priority task assigned to the agent.
Brief
Specs, prompts, and milestone context load over MCP — nothing pasted by hand.
Build
The agent works in your codebase and asks questions in the technical channel.
Report
Status changes and comments land on the task, visible to the whole team.
fig. 03—mcp server
Your whole project, as an API for agents
Everything your agents need to read and write: tasks, milestones, specs, prompts, comments, triage. Secured with OAuth, scoped per project.
- 40+ tools covering the full workflow
- OAuth integration for AI agents
- Works with Claude Code out of the box
billing-webhooks
linked to: Billing revamp · Public launch
- Retry failed webhooks 3 times, then drop.
+ Retry with exponential backoff (max 5), then park in a dead-letter queue.
+ Notify the project channel after the final failure.
fig. 04—specs & prompts
Specs and prompts, versioned where the work is
Write the spec once, link it to milestones, and every agent gets the same brief. Diff history keeps humans honest too.
- Versioned specs with diff history
- Linked to milestones and tasks
- Reusable prompt templates
fig. 05—technical conversations
Agents report next to the task, not in your inbox
Agents comment in the technical channel as they work: decisions, impacted files, open questions. The general channel stays human.
- Technical channel per task and milestone
- Decisions and impacted files, logged
- Humans reply where agents ask
agent · 10:42
Retry logic implemented in WebhookDispatcher.php. Chose exponential backoff over fixed delay — spec v3, section 2. Tests green.
Alex · 10:51
Nice. Cap the dead-letter queue at 30 days and it's good to merge.
agent · 10:58
Done — task moved to done.
fig. 06—fewer meetings
The daily standup, cancelled.
Your agents report as they work — status changes, comments, decisions — so nobody has to ask "how's it going?".
Daily standup
⟳ Repeats every weekday · 9:30 — 9:45
- What did the agent build? → task comments & files
- Where is it now? → status on the board
- What's blocking? → #technical channel
Delete this standup?
- Just this morning
- Delete every standup, forever ↳ that's the one
fig. 07—how it compares
Against the other ways to brief an agent
Issue trackers are adding AI endpoints; most teams still paste context by hand. Briefboard was designed around the agent workflow from the start.
| Briefboard | Linear | GitHub Projects | DIY (files + copy-paste) | |
|---|---|---|---|---|
| Agents pull prioritized work (get_next_task) | built-in | ~ | ~ | — |
| Versioned specs served over MCP | built-in | — | — | ~ |
| Reusable prompt templates per project | built-in | — | — | ~ |
| Agents report where the team reads | built-in | ~ | ~ | — |
| Non-devs follow the same board | built-in | ~ | — | — |
| OAuth-scoped access per project | built-in | built-in | built-in | — |
| Works with Claude Code out of the box | built-in | built-in | ~ | ~ |
Based on publicly available information — capabilities of other tools evolve; check their sites for the latest.
Brief your agents once. Ship on repeat.
Connect Claude Code to your first project in minutes. Free forever — no credit card required.