Free · MIT · Open source

Open Research

An open-source research agent. Ask a question; it plans searches, reads the most relevant pages, and answers with numbered citations you can check.

Get a CompanyFabric keyGitHub repo coming soon

Features

What it does

Answers with numbered citations

It plans searches, reads the most relevant pages, and every [n] points at a source a tool actually returned.

Guarded page reading

page_read fetches only public http(s) pages. It refuses localhost, internal names and private or metadata addresses, and re-checks every redirect.

One tool layer with approvals

Every tool call goes through one broker that applies the agent's policy, asks you before anything with side effects, and writes an audit row first.

Saved reports

Threads, messages, the audit log, approvals and reports live in one SQLite file.

Remix in one file

agent.config.ts holds the instructions, a policy per tool (allow, approve or deny) and the limits. The broker enforces them on the server, not the prompt.

One key, every model

Use with your CompanyFabric key

  1. 1

    Create an account and make a key on the keys page. New accounts get starter credit.

  2. 2

    Put it in the template's .env. companyfabric/auto picks a suitable model per request, and the exact charge of each turn comes back with the answer.

  3. 3

    Run it: pnpm dev for lite (Vercel works), or docker compose up for Hermes. With no key it runs a clearly labelled demo.

OPENAI_BASE_URL=https://api.companyfabric.com/v1
OPENAI_API_KEY=sk-cf-…
MODEL=companyfabric/auto

OpenAI directly, or a local server such as Ollama, LM Studio or vLLM, works too: any OpenAI-compatible endpoint with tool calling.

Runtimes

Lite or Hermes

Lite runs the agent loop in the Next.js server. Hermes runs each turn in a Hermes Agent container (v0.22.5, pinned by digest), which calls the app's tools back over MCP, so policies, approvals and the audit log are the same.

CapabilityLiteHermes
Web search with numbered citationsYesYes
Read web pages (SSRF-guarded)YesYes
Approval before side effectsYesYes
Audit row before every tool callYesYes
Save reportsYesYes
Memory across threadsHermes memory toolsetNoYes
Files and terminal in a sandboxEnable the file / terminal toolsets in hermes/config.yamlNoOff by default
Browser for JavaScript-heavy pagesHermes browser_exec; needs the terminal toolsetNoOff by default
Scheduled research (cron)Hermes cron works; this UI does not manage schedules yetNoNot in the UI yet
Runs on serverless hosting (Vercel)Hermes needs a long-running containerYesNo

What is proven

Status

Early. Checked in CI on every change: Install, typecheck, unit tests and production build; Hermes starts with the app's tools loaded (no model is called in CI).

  • implemented-untestedSearch through Parallel (keyless tier)
  • implemented-untestedSearch through Exa
  • implemented-untestedLive research eval (10 questions, citation checks)
  • implemented-untestedA Hermes turn with a real model
  • not yetCompanyFabric search: not released yet
  • not yetManaging scheduled research from the UI

Configuration

Environment

VariableDefaultWhat it is
OPENAI_BASE_URLhttps://api.companyfabric.com/v1Any OpenAI-compatible chat.completions endpoint with tool calling.
OPENAI_API_KEYnoneThe model key. With none set, the app runs a labelled DEMO.
MODELcompanyfabric/autoThe model id. companyfabric/auto picks a suitable model per request.
SEARCH_PROVIDERparallelparallel (free, rate-limited, no key) or exa. companyfabric is not released yet.
PARALLEL_API_KEYnoneLifts Parallel's free-tier limits.
EXA_API_KEYnoneNeeded when SEARCH_PROVIDER=exa.
COMPANYFABRIC_SEARCH_ENABLEDfalseCompanyFabric search is not released yet; leave it off.
RUNTIMEnonelite or hermes. Empty: hermes when HERMES_URL is set, lite otherwise.
HERMES_URLnoneSet for you by docker-compose.yml.
HERMES_API_KEYnoneHermes only. Generate with: openssl rand -hex 32
TOOL_TOKENnoneHermes only: lets Hermes call the app's tools at /api/mcp. Generate with: openssl rand -hex 32
DATA_DIR./dataWhere the SQLite database lives.

MIT

License

Open Research is MIT-licensed, © 2026 Creative Tim. Its UI is adapted from AI Elements (Apache-2.0) and shadcn/ui (MIT).

Not affiliated with any company whose AI products offer similar research features.

All templates →