Skip to main content
Glama

WARNING

Public beta. Standcast is being tested with its first teams. Expect rough edges and changes to behaviour, configuration and the API before 1.0. Report problems in GitHub issues or at support@standcast.app.

Each person runs /daily in their AI agent (Claude Code, Cursor, Codex, Claude Desktop, VS Code). Standcast drafts the daily from what they did on GitHub; the person edits it, adds today's plan and blockers, and confirms before anything is sent. At the time set in standcast.yml, Standcast posts one digest per team to Slack or Google Chat.

It is built for teams, not for watching people: no productivity metrics, no rankings, nothing sent on anyone's behalf.

Table of contents

Related MCP server: Slack Gitlab Digest Model Context Protocol MCP Server

Quick start

New here? The landing page has the overview and the beta request form. Agents can read llms.txt.

For admins

You need to own the GitHub organization (or personal account).

  1. Install the Standcast GitHub App on your organization.

  2. GitHub sends you to https://app.standcast.app/setup. Sign in. If your teams post to Slack, click Connect Slack. If they post to Google Chat, add the Standcast app to a space and paste the code it posts in the Google Chat step (guide).

  3. Create standcast.yml at the root of your private .github-private repository (or of .github), from the template on the setup page.

Step by step: Install.

For members

Install the CLI (macOS and Linux) and sign in:

curl -fsSL https://standcast.app/install.sh | sh
standcast login

The script downloads the latest release for your system, checks it against SHA256SUMS and installs it to ~/.local/bin/standcast. Windows binaries are on the releases page.

Then pick how you write your daily:

Client

Setup

Claude Code

/plugin marketplace add ImFelipeOliveira/standcast, then /plugin install standcast@standcast, then run /daily

Cursor, Codex, Claude Desktop, VS Code

Add the MCP server (standcast mcp): MCP guide

Terminal

standcast draft and standcast submit: CLI reference

Why Standcast

Once a team writes dailies with AI, three problems show up:

  • Loose dailies. Every message comes in a different format, in a different place.

  • No consolidated view. The manager reads scattered messages and has no single picture of the team.

  • Nobody orchestrates. No one tracks who sent the daily, and blockers sit until someone notices.

Standcast keeps the writing where people already work, inside their agent, and takes care of collecting, consolidating and delivering.

Features

  • Draft from GitHub activity. Pull requests opened, merged and reviewed, plus commits grouped by repository, since your last daily. Monday covers the weekend.

  • Works in any MCP client. A Claude Code plugin with the /daily skill, an MCP server for Cursor, Codex, Claude Desktop and VS Code, and a plain CLI for the terminal.

  • One digest per team. Posted to Slack or Google Chat at digest_at, in the team's time zone and only on workdays, with blockers first and the people who did not send listed at the end.

  • Late dailies. Sent after the digest, they are posted as a reply in the digest thread.

  • Reminders. At remind_at, people who have not sent their daily get one direct message, at most once a day, and never after the digest.

  • Blocker routing. A new blocker goes by direct message to whoever can help: the pull request's reviewers, then its CODEOWNERS, then the team managers.

  • Slack or Google Chat. Each team picks its chat in standcast.yml. On Google Chat, add the Standcast app to a space and link it with the code it posts; people link themselves by sending the app a direct message.

  • Linear cards. Connect one Linear workspace (read-only) and your own cards that started or finished appear in your draft. A blocker that links a card goes to the people who own the blocking card.

  • Suggested blockers, only for you. A pull request waiting too long for review, CI failing for a day or a card in progress for days shows up in your own draft. You confirm or ignore it; nobody else ever sees it.

  • Several teams, one daily. People in two teams write once and appear in both digests.

  • Configuration as code. A single standcast.yml in your private .github-private repository (or in .github). Every push and pull request that changes standcast.yml is validated by a Standcast config check run with the error on the right line, and an invalid push never breaks anything: the last valid version stays in effect.

  • Manager tools. Managers read their team's dailies, digests and blockers from their own AI agent.

  • Read-only API and webhooks. API keys (sck_…), an OpenAPI reference, and signed outgoing webhooks such as daily.submitted and blocker.opened.

  • Bilingual. English and Brazilian Portuguese across the CLI, chat messages, web pages and documentation.

How it works

flowchart LR
    subgraph Member["Each member"]
        A["AI agent<br/>/daily · MCP · CLI"]
    end
    subgraph SW["Standcast"]
        API["API"]
        W["Worker"]
        DB[("Postgres")]
    end
    GH["GitHub<br/>activity · standcast.yml"]
    SL["Slack or Google Chat<br/>team channel or space"]

    A -- "1. draft" --> API
    API -- "reads activity" --> GH
    A -- "2. confirmed daily" --> API
    API --> DB
    W -- "3. digest at digest_at" --> SL
    DB --> W

A day with remind_at: "09:30" and digest_at: "10:00":

Time

What happens

08:50

Alice runs /daily. Her agent fetches a draft of yesterday's GitHub work; she edits it, writes today and one blocker, and confirms. Her blocker links a pull request, so its reviewer gets a direct message.

09:30

Bob has not sent yet, so he gets a reminder.

10:00

Standcast posts the team digest to #daily-backend: blockers first, then each person's yesterday and today, then who did not send.

10:40

Bob sends a late daily. It is posted as a reply in the digest thread.

The full walkthrough is in How it works.

Configuration

All configuration lives in standcast.yml at the root of the .github-private repository or, when that one has no file, of the .github repository, on the default branch. Keep it in a private repository: it names the people of your teams. A minimal file:

teams:
  - name: backend
    timezone: America/Sao_Paulo
    remind_at: "09:30"
    digest_at: "10:00"
    managers: [alice]
    members: [alice, bob]
    chat:
      provider: slack
      channel: "#daily-backend"

Every key is documented in the configuration reference.

Privacy principles

  • Your code is never read. Standcast reads standcast.yml, team and organization membership, and the titles, links and dates of your own pull requests, reviews and commits. If you connect Linear, it also reads card titles, links, states and dates, never descriptions or comments.

  • Nothing is sent without your confirmation.

  • No rankings, scores or productivity metrics, in any plan, ever.

  • No AI model processes your data on the Standcast servers. The writing happens in your own agent.

  • Secrets are encrypted at rest with AES-256-GCM.

Details: Privacy.

Architecture

A Bun + TypeScript monorepo. Business rules live in use cases; HTTP routes, jobs, the CLI and the MCP server are thin layers on top of them.

apps/
├── api/            Hono HTTP API, OAuth, GitHub, Slack and Google Chat webhooks, /reference
├── worker/         pg-boss jobs: tick, digests, late dailies, webhook delivery
└── cli/            standcast CLI and MCP server (compiled to a single binary)
packages/
├── core/           pure domain: config, calendar, digest builders, i18n
├── orchestrator/   use cases: drafts, submissions, digests, org reads
├── db/             Drizzle schema, migrations and repositories (Postgres)
├── queue/          job queue port and pg-boss implementation
├── plugin/         Claude Code plugin with the /daily skill
└── adapters/
    ├── github/     GitHub App: activity, config, teams, OAuth
    ├── slack/      Slack: OAuth, digests, threads, direct messages
    └── google-chat/ Google Chat: app events, digests, threads, direct messages
docs/
├── site/           VitePress documentation (en and pt-BR), hosted on Cloudflare Pages
└── api/            OpenAPI guides and changelog

Dependency direction: apps → orchestrator → core, with db, queue and the adapters behind ports so use cases are tested without HTTP, Slack or GitHub.

Self-hosting

Standcast ships as one Docker image containing the api and the worker. APP (api or worker) selects the default command. The documentation site is not part of the image: it is a static site on Cloudflare Pages, see docs/cloudflare-pages.md.

docker build --build-arg APP=api -t standcast-api .
docker build --build-arg APP=worker -t standcast-worker .
  • The api runs the migrations on startup, so run a single api replica. Both services need Postgres.

  • Every variable is listed in .env.example.

  • Railway setup: docs/deploy.md.

  • Register the apps with docs/github-app.md (or the manifest), docs/slack-app-manifest.yml and docs/linear-app.md. LINEAR_CLIENT_ID and LINEAR_CLIENT_SECRET are required on the api and the worker.

  • Google Chat is optional: create the Chat app with docs/google-chat-app.md, then set GOOGLE_CHAT_SERVICE_ACCOUNT (the base64 of the service account JSON key) on the api and the worker, and GOOGLE_CHAT_PROJECT_NUMBER (the Google Cloud project number, api only, required with the key). The Chat app authentication audience must be Project number. Without the key, Google Chat is off.

Development

bun install
bun test                    # starts one disposable Postgres for the full suite
bun run lint                # Biome, CI-enforced
bun run typecheck           # tsc, CI-enforced

Tests require Docker, but no database setup: Testcontainers starts PostgreSQL 17 once, shares it across the whole run and removes it after the last test. Set TEST_DATABASE_URL to reuse an existing database instead. To run the api or worker locally, copy .env.example to .env and run docker compose up -d.

Documentation site, served at the root of its own domain (https://docs.standcast.app). STANDCAST_PUBLIC_URL sets the api address shown in the examples and the API reference link:

SUPPORT_EMAIL=you@example.com bun run docs:dev
SUPPORT_EMAIL=you@example.com bun run docs:build
bun run check-docs

Release automation

  • CI (lint, typecheck, tests against Postgres, docs build) runs on every pull request and on pushes to dev and main.

  • Feature branches merge into dev; dev merges into main when a release is ready.

  • release-please keeps a release PR open on main from Conventional Commits. Merging it tags the release and uploads the CLI binaries and SHA256SUMS.

Contributing

Contributions are welcome. Read CONTRIBUTING.md for commit conventions, releases, local setup and code style before opening a pull request.

Roadmap

  • Drafts from GitHub activity, through the skill, the MCP server and the CLI

  • Team digest and late dailies in Slack

  • Manager tools, read-only API and outgoing webhooks

  • Reminder for people who have not sent their daily

  • Blocker routing to reviewers, CODEOWNERS or managers

  • Suggested blockers (pull request without review, failing CI), visible only to you

  • standcast.yml validation as a GitHub check run

  • Linear integration: cards in the draft, card blocker routing and stale card suggestions

  • Google Chat: digests, threads and direct messages through the Standcast app

  • Jira, then ClickUp, Asana and GitHub Projects

  • Microsoft Teams and Discord

Sponsors

Orange Labs supports Standcast. Financial support for the project goes through GitHub Sponsors.

Support

Questions and problems: support@standcast.app.

License

MIT © 2026 Felipe Oliveira

Available Tools

8 tools
standcast_decide_suggestionConfirm or ignore a suggested blockerA
Idempotent

Records the person's answer to a suggested blocker from their own draft. Call it only after asking the person whether the suggestion is a real blocker. Use decision ignored when they say no. When they say yes, add it to the blockers in their own words with its ref; submitting the daily confirms it, and decision confirmed records the answer right away. Calling with decision confirmed is optional, because submitting a daily with a blocker that carries the suggestion's ref confirms it. Never decide on the person's behalf.

Inputs: id (required), decision (required), org. Returns: id, decision

standcast_get_digestGet a team digestA
Read-onlyIdempotent

Returns the digest delivery of a team on one date, with who had not sent the daily, or null when there is none. Limited to the last 90 days.

Inputs: date (required), team (required), org. Returns: digest, untrusted

standcast_get_draftGet the daily draftA
Read-onlyIdempotent

Builds the draft of yesterday's work from the person's GitHub and Linear activity and, unless includeSessions is false, the person's local AI sessions (Claude Code), read on this machine and never sent to Standcast. work groups activity and sessions into work items: build yesterday as a short summary of at most 5 lines, grouping related work items by project or theme. plan (the previous daily's plan against what moved) and pending (the person's open pull requests, review requests and cards) are only for the person; plan is never copied into the daily. warnings names sources that were unavailable. Use it as a starting point: the person decides what goes in the daily.

Inputs: org, includeSessions. Returns: org, date, plan, work, +7 more

standcast_list_blockersList team blockersA
Read-onlyIdempotent

Lists the blockers confirmed by the members of a team, newest first, paginated. Without from and to, covers the last 90 days.

Inputs: team (required), to, org, from, +1 more. Returns: data, untrusted, next_cursor

standcast_list_dailiesList team dailiesA
Read-onlyIdempotent

Lists the confirmed dailies of a team between two dates, newest first, paginated. Limited to the last 90 days.

Inputs: to (required), from (required), team (required), org, +1 more. Returns: data, untrusted, next_cursor

standcast_list_teamsList the teams I can readA
Read-onlyIdempotent

Lists the teams the person manages, or every team for an organization admin. Only these teams can be read with the other manager tools.

Inputs: org. Returns: data, untrusted, next_cursor

standcast_submit_dailySubmit the dailyA
Idempotent

Submits the person's daily to their teams. Before calling, you MUST show the final daily to the person and get their explicit confirmation. When the client supports elicitation, Standcast also asks the person directly and sends nothing unless they accept. Otherwise, call with confirmed: true only after the person explicitly confirmed. Send each item exactly as the person wrote or approved it: never translate or rewrite it. Submitting again on the same day replaces the previous daily.

Inputs: today (required), blockers (required), yesterday (required), org, +2 more. Returns: date, late, teams, confirmation

standcast_whoamiWho am IA
Read-onlyIdempotent

Returns the GitHub login signed in to Standcast and the organizations.

Inputs: no named parameters. Returns: orgs, login

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4/5.0

    Scored across 8 tools

    Disambiguation4/5

    Most tools occupy clearly separate slots: read-only team listings (list_teams, list_dailies, list_blockers, get_digest), a personal draft builder (get_draft), an identity tool (whoami), and two write paths (submit_daily, decide_suggestion). The one fuzzy area is decide_suggestion versus submit_daily, since the descriptions explain that submitting a daily can implicitly confirm a blocker suggestion—an agent has to read carefully to know it may skip decide_suggestion. The long, cross-referencing description mitigates this rather than creating real misselection.

    Naming Consistency4/5

    Seven of eight tools follow a uniform snake_case standcast_verb_noun pattern (get_digest, get_draft, list_*, submit_daily, decide_suggestion) with a shared server prefix. whoami breaks the verb_noun convention, though it is a recognizable idiom. Otherwise the naming is fully predictable.

    Tool Count5/5

    Eight tools is well within a comfortable range, and each maps to a distinct step in the daily-standup workflow: orient (whoami, list_teams), gather (get_draft, get_digest, list_dailies, list_blockers), then act (submit_daily, decide_suggestion). No tool looks like padding, and none is missing so critically that the set feels thin.

    Completeness4/5

    The write side of the workflow is fully covered: drafting, blocker decisions, submission, and a complete read surface for teams, dailies, blockers, and digests. Minor gaps remain: there is no tool to fetch a single daily by id or date, no way to create or resolve a blocker outside a daily submission, and no team-membership lookup to resolve names returned by list_dailies/list_blockers. These are workable around but leave the surface slightly short of complete.

    Maintenance

    ActivityNo data
    ResponsivenessOutdated data

    Related MCP Connectors

    Related MCP Servers