Updated today
by GetStream
Vision-Agents
Quickly build text, voice and video agents. Fast, Go based Agent backend with SDKs for every language. Sensible defaults for search, knowledge, skills, telephony, sandboxes, memory and persistence. Sovereign AI. Take control of your AI.
8.2k GitHub stars (8,216)
- 30 skills (30 skills found in its SKILL.md files)
- 701 forks (701 forks)
- Today last update (Last updated today)
Install
Skills CLI
npx skills add GetStream/Vision-Agents
Asks which skills and agents to install. Just one: npx skills add GetStream/Vision-Agents --skill pr-review — add -g for every project.
Claude Code
git clone https://github.com/GetStream/Vision-Agentscp -r "Vision-Agents/.agents/skills/pr-review" ~/.claude/skills/
Copies the pr-review skill into ~/.claude/skills/ for every project (use .claude/skills/ for one project). Repeat for the other skills listed below.
Plugin
/plugin marketplace add GetStream/Vision-Agents
Run inside Claude Code if the repository ships a plugin marketplace, then pick plugins from /plugin.
Skills can run code on your machine — read the SKILL.md first.
What it does
In the author's own words, from each skill's SKILL.md — showing the first 30 of 55 files.
-
pr-review
Reviews a GitHub pull request in this repo at its exact head commit, in a detached git worktree when the current checkout differs, against the PR description, existing review threads, CI, and this repo's rules (OpenAPI spec and client regeneration, x-client-accessible operations, SDK order, migrations). Use when the user gives a PR number (123, #123), a github.com/.../pull/123 link, or asks to re…
.agents/skills/ on GitHub (opens in a new tab)pr-review/ SKILL.md -
typesafe-ai
Build AI-powered software with TypeSafe: small units of AI intelligence you can use like programming primitives. Its System One models, including Jev, turn natural language and application state into typed judgments and probabilities that code can combine. Use when a feature needs programmable common sense, when brainstorming what AI could make possible in an app, or when an LLM prompt-and-parse…
.agents/skills/ on GitHub (opens in a new tab)typesafe-ai/ SKILL.md -
vgpu
Build and optimize WebGPU apps with the vgpu package and entrypoints vgpu/node, vgpu/mock, vgpu/scene, and vgpu/client. Use @vgpu/render/inspect, /utils, /edit, and /perf only as slim tooling subpaths. Bundles performance guides and the API reference; load one doc at a time.
.agents/skills/ on GitHub (opens in a new tab)vgpu/ SKILL.md -
accelerate-sales-engineer
Build a short Stream sales-engineering demo for a customer call: one Accelerate agent, their branding, a streaming web chat UI, and a tunnel link to review. Use this whenever someone asks for a customer demo, a POC, a call demo, a sales-engineering build, a tailored agent for a prospect, or to make something to show a company from a name and URL. Also use it when they mention the Stream CLI, Stre…
.claude/skills/ on GitHub (opens in a new tab)accelerate-sales-engineer/ SKILL.md -
add-connector
How to add a connector - a custom MCP server for one app (no code), or a new built-in in the router's catalog (a manifest YAML with sources, a consent test and a PR). Read before you add a provider, change a built-in manifest, or bump its revision.
.claude/skills/ on GitHub (opens in a new tab)add-connector/ SKILL.md -
check_prices
Check every price the router bills with against what each vendor charges today, and prepare the price changes for each one that is wrong. Use when a vendor announced new prices, before a release, or when asked whether our prices are right.
.claude/skills/ on GitHub (opens in a new tab)check_prices/ SKILL.md -
cli
How the Stream CLI (getstream) reaches the agents backend, and how to point it at a local router or the hosted one. Read before running getstream agents commands or debugging why one fails.
.claude/skills/ on GitHub (opens in a new tab)cli/ SKILL.md -
commit
Create a git commit for the Vision-Agents repo.
.claude/skills/ on GitHub (opens in a new tab)commit/ SKILL.md -
connectors
Walks a developer through connectors end to end - the terms, turning them on in the router, connecting an account, giving an agent a connector's tools (one app account or each user's own), a channel such as the Slack bot, and what to check when something stays silent. Read before you connect a connector to an agent, set one up locally or on staging, or debug one.
.claude/skills/ on GitHub (opens in a new tab)connectors/ SKILL.md -
dashboard
How to work on the Volt dashboard (/Users/thierry/workspace/volt-dashboard), Stream's admin SPA, and its Agents section that talks to the acceleration router. Read before touching Volt or wiring a router change into it.
.claude/skills/ on GitHub (opens in a new tab)dashboard/ SKILL.md -
decision_model
What a decision model is in this repo, the contract one implements, the System One protocol most vendors here speak, and how to add a vendor or a model to the decision_model router. Read before adding a decision model, a decision model vendor or a guardrail backend.
.claude/skills/ on GitHub (opens in a new tab)decision_model/ SKILL.md -
deploy
Ship acceleration/cmd/router to Stream's hosted environments. Read before deploying the router or asking why the hosted one behaves differently.
.claude/skills/ on GitHub (opens in a new tab)deploy/ SKILL.md -
docs
How to edit the agents docs on getstream.io. Use when writing or updating agents documentation pages, sidebars or the SDK switcher.
.claude/skills/ on GitHub (opens in a new tab)docs/ SKILL.md -
generative-ui
How an agent's interactive UI (OpenUI, A2UI) travels from the model to the apps - the agent.yaml `ui:` key and the prompt it compiles, the router's streaming splitter, the `ui_delta`/`ui_completed` socket frames, what is kept in Stream Chat, and LiveUI in the client SDKs. Read before changing any of them or debugging a card that is missing, stuck or jumps.
.claude/skills/ on GitHub (opens in a new tab)generative-ui/ SKILL.md -
go-testing
How Go tests are written in acceleration/ and sdks/go. Read before adding a test or a test suite.
.claude/skills/ on GitHub (opens in a new tab)go-testing/ SKILL.md -
go
How Go is written in this repo - errors, logging, spans, context, goroutines, comments. Read before writing or changing Go code.
.claude/skills/ on GitHub (opens in a new tab)go/ SKILL.md -
good_morning
Pull and merge the latest code across the workspace repos (chat, Vision-Agents, artemis-impl, getstream.io, volt-dashboard). Use when the user says good morning or asks to sync their repos. Checks the team Slack channel and summarises new messages from the last day.
.claude/skills/ on GitHub (opens in a new tab)good_morning/ SKILL.md -
integrate_new_models
Find the models worth adding with the new_models skill, then build support for each one in its own branch and open a draft PR per model.
.claude/skills/ on GitHub (opens in a new tab)integrate_new_models/ SKILL.md -
llm-interface
Consider changes to the llm standardization/interface
.claude/skills/ on GitHub (opens in a new tab)llm-interface/ SKILL.md -
loop_performance_bench
Run the router performance loop - profile locally, fix the top cost, cut needless API, SQL and Redis calls, deploy to the benchmark cluster, ramp, record agents held and $/voice-minute, repeat ten times. Use when asked to run the performance loop, to improve how many agents one router holds, or to lower the hosting cost per voice minute.
.claude/skills/ on GitHub (opens in a new tab)loop_performance_bench/ SKILL.md -
migrate-to-accelerate
Migrate a voice or text agent from LiveKit, Retell, Vapi, Pipecat, Decagon, Sierra AI, GPT Live, or Deepgram Agents to a Stream accelerate agent. Use when the user asks to migrate, port, convert, or rewrite an agent from one of those platforms, or hands over a competitor agent's code, dashboard export, or API object.
.claude/skills/ on GitHub (opens in a new tab)migrate-to-accelerate/ SKILL.md -
new_models
Find LLM, TTS, STT and STS models worth adding to the router by searching X and Artificial Analysis, then diffing against what we already route.
.claude/skills/ on GitHub (opens in a new tab)new_models/ SKILL.md -
pagination
How list endpoints page through results. Read before adding a list endpoint or a limit parameter.
.claude/skills/ on GitHub (opens in a new tab)pagination/ SKILL.md -
parallel-agents
Rules for running several coding agents on this repo at once (PR authors, reviewers, fixers, rebases), so they do not break each other's runs and do not burn tokens on repeated work. Read before you start more than one agent that writes code, runs integration tests or opens PRs.
.claude/skills/ on GitHub (opens in a new tab)parallel-agents/ SKILL.md -
performance_benchmark
Run the voice-agent load benchmark in performance_benchmark/ to find how many cascaded agents one router holds, and read or compare its reports. Use when asked how many concurrent agents we support, to load-test or stress the router, to measure a performance change, or to explain a perfbench run.
.claude/skills/ on GitHub (opens in a new tab)performance_benchmark/ SKILL.md -
phonecheck
Test inbound or outbound phone calls through a local or remote router. Asks what to test, sets up numbers, hooks and the person side, places the calls with phonecheck, reports where each hop ran (agent SFU, router address, Stream SIP host, SFUs used, person side) and how long it took, then puts everything back. Use when asked to test phone calls, run a phone call check, see which servers a call u…
.claude/skills/ on GitHub (opens in a new tab)phonecheck/ SKILL.md -
pr
Create a draft pull request for the Vision-Agents repo using gh CLI.
.claude/skills/ on GitHub (opens in a new tab)pr/ SKILL.md -
pricing
How to make new billable work (a model, vendor, modality or phone feature) report usage and get priced correctly. Read before adding anything the router bills, or when work shows up unpriced.
.claude/skills/ on GitHub (opens in a new tab)pricing/ SKILL.md -
prospect-competitor-research
Research a prospect's top 5 competitors and write a dated, sourced report: press releases, latest blog posts, changelogs, new features, where each competitor is ahead, and who is shipping AI features and how those features surface in the product. Use when someone asks for competitive research, a competitor analysis, a market scan, or call prep for a prospect or customer, given a company name and…
.claude/skills/ on GitHub (opens in a new tab)prospect-competitor-research/ SKILL.md -
provider_audit
Audit what is left on every vendor account the product pays for — prepaid balance, credits, character or minute quota, spend this month — read-only, and say which account is about to run out. Use when asked how much money is left, to audit providers, for a vendor's balance, credits or quota, or whether any account is about to run out.
.claude/skills/ on GitHub (opens in a new tab)provider_audit/ SKILL.md
No skill matches that filter.
Review before you install
A skill is a set of instructions and scripts your agent will follow, and it can run code on your machine. Read the SKILL.md and any scripts first — this listing is not a security review.