Repository navigation
Conversation
📝 Summary
Merge Risk:
 |
| Check name | Status | Explanation |
|---|---|---|
| Title check |  |
The title clearly identifies the main change: adding the 2026 Company AI Brain blog guide. |
| Description check | |
The description directly matches the changeset and lists the guide, route, formatting checks, FAQ content, and supporting images. |
| Docstring Coverage | |
No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0… |
| Linked Issues check | |
Check skipped because no linked issues were found for this pull request. |
| Out of Scope Changes check | |
Check skipped because no linked issues were found for this pull request. |
Full details: Docstring Coverage
-
Explanation
No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0 files. (1 skipped: 1 unsupported.)
Finishing Touches
-
🧪 Generate unit tests (beta)
-
- Commit to this branch
- Create a new PR
-
- Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.
Comment @coderabbitai help to get the list of available commands.
Website previewPreview URL: https://pr-7769-agenta-website-preview.mahmoud-637.workers.dev Built from |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 5
-
-
- Source mode: Auto
- Repository: Agenta-AI/agenta
- Branch: agent/how-to-build-a-company-ai-brain
- File: .coderabbit.yaml
- Commit: 51982e2
- Layers: Repository YAML (base), Organization UI (inherited)
- Review profile: CHILL
- Plan: Advanced
- Run ID:
c4ef911d-4ebb-4a28-990a-535b950ebd05
-
web/website/public/blog/how-to-build-a-company-ai-brain/agenta.webpweb/website/public/blog/how-to-build-a-company-ai-brain/company-ai-brain-architecture.webpweb/website/public/blog/how-to-build-a-company-ai-brain/hero.webpweb/website/public/blog/how-to-build-a-company-ai-brain/passive-rag-vs-active-company-brain.webpweb/website/src/content/posts/how-to-build-a-company-ai-brain.mdx
Included review availability: This review used your included allowance. Your plan provides up to 8 included reviews per hour; 7 remain after this review.
-
| | **[Pi](https://agenta.ai/docs/concepts/harnesses-and-models)** | Multi-provider models: OpenAI, Anthropic, Google Gemini, Mistral, Groq, MiniMax, Together AI, OpenRouter, xAI, and any OpenAI-compatible endpoint | Teams that want maximum model flexibility across providers and custom gateways on a single harness | | ||
| | **[Codex](https://github.com/openai/codex)** ([docs](https://agenta.ai/docs/changelog/codex-harness)) | OpenAI's Codex-optimized models (including GPT-5.6 Sol, Terra, and Luna, plus GPT-5.5 and GPT-5.2) directly via OpenAI | Fast software engineering, repository analysis, data scripts, and structured tool execution | | ||
|
|
||
| Because the harness, model, and credential are decoupled, switching from one model provider to another, or moving from direct API keys to AWS Bedrock, is a two-click configuration change. Your company brain's instructions, skills, shared files, and permissions remain completely untouched. |
There was a problem hiding this comment.
🎯 Functional Correctness |
Qualify model and gateway switching by harness.
The article says moving to AWS Bedrock is a configuration change without noting the harness limit. Agenta documents that Codex reaches OpenAI directly and cannot use Bedrock, Vertex AI, or a custom gateway. A team that chooses Codex cannot make that move without changing harnesses. State this constraint where the article promises portability. (agenta.ai)
|
|
||
| Next, decide where your company brain runs and how model access is governed: | ||
|
|
||
| - **Self-Hosted (MIT Open Source):** If your security posture requires keeping all files, prompts, and run logs inside your own VPC, deploy Agenta via Docker Compose. You get unlimited users, projects, agents, and automations with role-based access control and SSO at zero software licensing cost. |
There was a problem hiding this comment.
🔒 Security & Privacy |
Do not equate self-hosting with keeping prompts inside the VPC.
Docker Compose can keep Agenta’s application and storage on company infrastructure, but the harness sends conversations to the selected model. The article later recommends provider API keys, and Agenta supports direct external model connections. A team following this sentence could send sensitive prompts outside its VPC despite selecting self-hosting for data residency. Specify that the model endpoint, storage, and connected services must also meet the residency requirement. (agenta.ai)
| Next, decide where your company brain runs and how model access is governed: | ||
|
|
||
| - **Self-Hosted (MIT Open Source):** If your security posture requires keeping all files, prompts, and run logs inside your own VPC, deploy Agenta via Docker Compose. You get unlimited users, projects, agents, and automations with role-based access control and SSO at zero software licensing cost. | ||
| - **Agenta Cloud:** If you want zero infrastructure maintenance without paying a per-seat tax, start on Agenta Cloud Hobby (Free) or Starter ($29/month), which includes **unlimited team members**, unlimited projects, and shared team credentials billed purely on runtime and model credits. |
There was a problem hiding this comment.
🎯 Functional Correctness |
Separate Hobby limits from Starter benefits.
This sentence groups Hobby and Starter under unlimited team members and projects. Hobby permits two team members and one project; those unlimited allowances start with Starter. State the Hobby limits here so a team does not choose the free plan for an unsupported rollout. (agenta.ai)
|
|
||
| Review the tools attached to your coworker and match permissions to how the agent will be triggered: | ||
|
|
||
| - **For interactive coworkers (Chat, Slack, Web Embeds):** Use the default safety posture where read actions (`search`, `fetch`, `list`, file reads) run automatically, and external write or send actions (`send_email`, `post_message`, `update_crm_record`) pause and ask for human approval. In a live chat or Slack thread, clicking **Approve** takes one second and eliminates the risk of an unwanted external send. |
There was a problem hiding this comment.
🔒 Security & Privacy |
Do not treat embed-visitor approval as operator approval.
For a web embed, Agenta shows an approval request to the visitor, who can approve it. The article also proposes embeds for customer portals. If that agent has a privileged external-write tool set to ask, a customer can approve their own requested action; this setting does not provide staff review. Recommend denying privileged writes on visitor-facing agents or routing them to a separate operator-reviewed workflow. (agenta.ai)
| | **Product & Operations** | Roadmap priorities, customer interview summaries, weekly KPI definitions, open-items tracker | Weekly metric synthesis, feedback clustering, SOP drafting, vendor comparison | Monday morning cron schedule posting a comparison digest to `#ops-weekly` | | ||
| | **Engineering** | Architecture docs, API conventions, release checklists, incident postmortems | PR review helpers, migration verification, trace inspection, docs sync | `Claude Code` or `Codex` harness in isolated sandbox containers with GitHub integration | | ||
|
|
||
| Notice the compounding effect: when Product updates `product-facts.md` after shipping a new capability, Marketing's content agent, Sales's battlecard agent, and Support's triage agent immediately work from the updated truth on their very next turn. |
There was a problem hiding this comment.
🗄️ Data Integrity & Integration |
Explain how updates reach different agents.
Agenta documents persistent files as belonging to one agent and being available across that agent’s sessions. Updating Product’s product-facts.md therefore does not, by itself, update the separate Marketing, Sales, and Support agents named here. This breaks the article’s central cross-department example. Describe a shared external source or an explicit synchronization step, or use one agent for the example. (agenta.ai)
Summary
Adds the SEO pillar guide How to Build a Company AI Brain: The Ultimate Guide for Teams & CTOs (2026) (
/blog/how-to-build-a-company-ai-brain).Included in this PR
web/website/src/content/posts/how-to-build-a-company-ai-brain.mdx(Prettier-formatted, zero em dashes, no duplicate lead description, verified Agenta docs & pricing, 7-item FAQ schema + visible section)web/website/public/blog/how-to-build-a-company-ai-brain/hero.webp(1600x900Identity v3.0 cover withfal.aimotif artwork)web/website/public/blog/how-to-build-a-company-ai-brain/passive-rag-vs-active-company-brain.webp(1600x900brand comparison diagram)web/website/public/blog/how-to-build-a-company-ai-brain/company-ai-brain-architecture.webp(1600x9005-layer architecture diagram)web/website/public/blog/how-to-build-a-company-ai-brain/agenta.webp(1600pxwide Agenta AI coworker workspace screenshot)