# Getting Started with Claude Code: A Setup Guide for Bid Managers
Install Claude Code from the desktop app, set up a folder for each tender with a bid brief, add a free skill and run a first go/no-go. No coding needed.
Getting started with Claude Code takes a bid manager four steps: open the Code tab in the Claude desktop app, point it at a folder that holds one tender and a short bid brief, add a free skill such as AI Go/No-Go, and give Claude one task. You don't need to know how to code. You type instructions in plain English, and in Manual mode Claude asks before it changes a file.
Claude Code comes with Claude's paid plans; the Free plan does not include it (Anthropic's [pricing page](https://claude.com/pricing), November 2026). This guide covers setup and a first task, and [the full Claude workflow for a bid](/blog/claude-for-rfp) shows where Claude Code fits among the other Claude tools.
## Do You Need to Know How to Code to Use Claude Code?
No. You type plain English into the Code tab of the Claude desktop app, which includes Claude Code, so there is no terminal or Node.js install to manage (Anthropic's [desktop quickstart](https://code.claude.com/docs/en/desktop-quickstart), November 2026).
Claude Code is an assistant that reads, writes and changes the files in a folder you choose, with your permission. Anthropic's [terminal guide](https://code.claude.com/docs/en/terminal-guide) says "You can use Claude Code even if you've never used a terminal before" (November 2026). For how it differs from chat, projects and Cowork, see [what skills, projects and Cowork are](/blog/claude-skills-vs-projects-for-rfps).
Three ideas are new, and none of them is code:
- **Folders.** Claude Code works inside the folder you open, so where you put the tender files decides what Claude reads.
- **Permission prompts.** Claude asks before it changes a file or runs a command, and you approve or decline each one.
- **The / menu.** Typing / in the prompt box lists the skills and commands you can call.
### What a Bid Manager Can Do with Claude Code
Claude Code suits bid jobs that start from files and end in a report or a draft. Four make good first tasks:
| Bid job | Skill (from AutoRFP.ai's free library) | What you get | What you check |
| -------------------------------- | ---------------------------------------- | -------------------------------------------------------------------------------------- | ------------------------------------------------------------------- |
| Qualify the tender | AI Go/No-Go | A scored recommendation with sources, deal-breakers and open questions | The deal-breakers, and the open questions only your team can answer |
| Build a compliance matrix | Compliance Matrix Builder | Every requirement with its source, exact wording, category and whether it is mandatory | Requirements it missed, especially in embedded files |
| Check a draft for contradictions | RFP Contradiction Checker | Conflicts inside the draft and against your source content | That each conflict is real |
| Draft answers | None needed: a bid brief with a gap rule | First drafts, with GAP wherever the files are silent | Send each GAP to the person who holds the answer |
## Getting Started with Claude Code Without the Terminal
The simplest install for a non-engineer is the Claude desktop app, which includes Claude Code and opens it on any folder you choose through its Code tab.
Before you start, you need three things (all as of November 2026):
- **A paid plan.** Pro, Max, Ultra, Team or Enterprise, per Anthropic's pricing page.
- **A Mac or Windows computer.** Linux is in beta.
- **An empty folder for the tender.** The next section fills it.
Then follow Anthropic's desktop quickstart:
1. **Install the app.** Download the Claude desktop app for macOS or Windows and run the installer.
2. **Sign in** with your Claude account.
3. **Open the Code tab.** Click **Code** at the top of the sidebar.
4. **Open the tender folder.** In the row under the prompt box, open the environment selector and pick the top entry under **Local**. Then click the folder control beside it, choose **Open folder…** and select your tender folder.
5. **Pick a permission mode** in the same row before you type the first task.
Anthropic lists four permission modes: Auto, Manual, Accept edits and Plan (November 2026). Start in **Manual**, where Claude asks before each edit or command, or **Plan**, where it proposes an approach without editing anything. Move to **Accept edits** once you trust the folder setup.
Other routes exist, per Anthropic's [Claude Code overview](https://code.claude.com/docs/en/overview) (November 2026): a one-command terminal install, extensions for VS Code and JetBrains, and [Claude Code on the web](https://code.claude.com/docs/en/claude-code-on-the-web), which runs in Anthropic's cloud. Check with IT before tender files go to the cloud.
## Set Up a Folder for Each Tender with a Bid Brief
Claude Code works on one folder at a time, so give each tender its own folder, with a bid brief file named CLAUDE.md that Claude reads at the start of every session. This layout works for most tenders:
```text
tender-name/
├── CLAUDE.md the bid brief, read at the start of every session
├── tender/ the issuer's files, as downloaded
├── company/ current company files only
├── outputs/ everything Claude writes
└── .claude/
└── skills/ skills for this tender, if you install them here
```
1. **Download the whole pack into `tender/`.** Include the Q&A and every attachment. A mandatory requirement that only appears in the Q&A, or a security workbook embedded in the RFP's Word file, is still part of the tender.
2. **Put only current files in `company/`.** Take out superseded policies and old drafts, so every file Claude can quote is one you would stand behind.
3. **Write the bid brief.** Copy the template below into CLAUDE.md and fill in the brackets.
4. **Keep the originals somewhere else.** Anthropic's advice on undoing changes (in its [first-day guide](https://support.claude.com/en/articles/14552382-your-first-day-in-claude-code), November 2026) relies on git, a version control tool that tender folders rarely use, so let Claude work on a copy.
Write the bid brief the way you would brief a new bid writer:
```text
Bid brief: [buyer], [RFP name and number]
Bidder: [your company]
Deadline: [date, time and method, as the RFP states them]
Files
- The issuer's files are in tender/. The Q&A is part of the tender:
read it with the RFP.
- Our current company files are in company/.
Rules
- If the files in tender/ and company/ do not support an answer, or part
of one, write GAP and say what is missing and who in the company would
hold it. Do not fill a gap with what a company like ours usually does.
- Quote approved answers word for word.
- Cite the tender section or Q&A number behind every finding.
- Save outputs in outputs/ with the date in the file name.
- Never send, upload or submit anything to the issuer.
```
The gap rule earns its line. Without one, Claude can mark a capability your files never mention as met; [what Claude did with and without a gap rule](/blog/claude-ai-strengths-and-weaknesses-for-rfps) shows the test.
Anthropic's Help Center calls setting up a CLAUDE.md file, which the `/init` command writes, "the single highest-value setup step you can take" (November 2026). It is built for code projects, so for a tender, write the brief yourself or ask Claude to draft it from the template.
Side note: Your plan's data terms apply to whatever Claude reads, locally or on the web. Before a confidential tender goes in, read [what to clear with IT before confidential tender files go into Claude](/blog/is-claude-safe-for-rfps), from training settings to instructions hidden in attachments.
## How Do You Install a Skill in Claude Code?
Install a skill in Claude Code by saving its folder in the tender folder or uploading it to claude.ai (the table shows where each route loads), then call it from the / menu.
1. **Get the skill.** Download AI Go/No-Go from [AutoRFP.ai's free skills library](/skills), free and unlocked with an email address. It comes as a `.skill` file: a zip that holds the skill folder.
2. **Check the version.** This guide uses version 1.1 (November 9, 2026). A copy downloaded earlier is version 1.0, which gave no formula for its score, so download it again. In version 1.1, Step 5 of SKILL.md points to "How to Calculate the Score" in `references/go-no-go-framework.md`.
3. **Install it** by one of the three routes in the table below.
4. **Call it.** Type / in the prompt box and pick the skill, or ask for the task in plain words: "Run a go/no-go on this tender."
| Route | What you do | Where the skill loads |
| ---------------------------- | ------------------------------------------------------------------------------------------------------------------ | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Upload it to claude.ai | Open **Customize**, then **Skills**, and upload the `.skill` file | Cowork, cloud and terminal sessions signed in with that account (Anthropic's [skills docs](https://code.claude.com/docs/en/skills), November 2026), and in our test a local Code-tab session |
| Save it in the tender folder | Unzip the `.skill` file into `.claude/skills/`, so the path ends `.claude/skills/ai-go-no-go-agent-skill/SKILL.md` | That folder only, as a project skill that travels with it |
| Save it for every folder | Unzip it into `~/.claude/skills/` as a personal skill | Every folder on your computer, but not Cowork or cloud sessions (Anthropic, November 2026) |
In our test on November 10, 2026, skills from our claude.ai account appeared in a local Code-tab session under the `anthropic-skills` prefix, so the go/no-go skill showed as `anthropic-skills:ai-go-no-go`. If yours does not appear in the / menu, save it in the tender folder instead.
Two tips if you have never done this:
- **Show hidden folders.** Folders whose names start with a dot are hidden in the Mac Finder until you press Command, Shift and full stop together.
- **Let Claude do it.** Ask: "Create .claude/skills in this folder and unzip ai-go-no-go-agent-skill.skill from my Downloads folder into it." Approve the command when Claude asks.
Anthropic's docs say the / menu shows the name set inside the skill, which for this one is `ai-go-no-go`, and that the folder name calls it too. Claude also loads a skill on its own when a request matches the skill's description (November 2026).
The nested `rfp-shredder/` folder inside the package will not appear as a second skill. And a skill is plain text that tells Claude what to do, so read its files before you install it: Anthropic advises using skills only from sources you trust ([Use skills in Claude](https://support.claude.com/en/articles/12512180-use-skills-in-claude), November 2026).
## Run a Go/No-Go as Your First Task
A go/no-go is a good first task in Claude Code because it only reads the tender, it ends in one report, and every finding in that report should point back to a section you can check.
1. **Ask Claude to list what it received.** Do it before any analysis, so a missing, cut-off or unreadable file cannot skew the result.
```text
Before we start, list every file you have received. For each file, give its name, what it contains and roughly how long it is, and say whether any part of it was unreadable, cut off or looked incomplete. Do not start any analysis yet.
```
2. **Run the skill in the same session,** so Claude works from the files it has just listed. Expect questions: the skill reads the company summary back for you to confirm, and marks anything the files cannot answer as REQUIRES USER INPUT. Answer what you know, or tell Claude to work from the files and say where they are silent.
```text
Use the AI Go/No-Go skill to assess whether we should bid on this tender. Our company profile is in company/: use those files as the company profile.
```
3. **Read the report.** Version 1.1 shows a scorecard with the arithmetic for each category, the overall percentage (with a range when inputs are missing), any deal-breakers, the items that need your input and next steps. Its decision is Strong Go, Conditional Go, Executive Review or No-Go, with No-Go below 50%. A deal-breaker, a mandatory requirement you cannot meet, gives No-Go whatever the percentage.
4. **Check the report before you act on it.** Treat it as the start of the bid decision: work through the checks below, and have a person answer every open item.
Plan it as a task you start and come back to, and remember that Claude Code shares usage limits with Claude chat (Anthropic's Help Center [article on Claude Code plans](https://support.claude.com/en/articles/11145838-use-claude-code-with-your-pro-max-or-ultra-plan), November 2026).
## What a First Go/No-Go Caught on a Public Tender
We ran the file check and the go/no-go on a public state agency software RFP, [State of Maine RFP# 202506081](https://www.maine.gov/dafs/bbm/procurementservices/vendors/rfps/rfp-archives) (issued July 15, 2025, closed August 8, 2025), for Tallowfield Systems, a fictional bidder, and submitted nothing. Each step ran once in Claude Code (Opus 5.5) on text extracted from the tender's Word files, outside the desktop Code tab: the file check on November 6, 2026 (about 7 minutes) and the go/no-go, on skill version 1.1, on November 9, 2026 (about 10 minutes).
In one run each:
- **The file check listed all 11 files** by exact name, with line counts that matched, and flagged problems before any analysis, among them a cut-off workbook label, a Q&A answer calling a 14-control tab blank, a pricing file ending on an open question, a superseded 2024 security policy and referenced documents that were missing. An independent scorer found nothing it had missed.
- **The go/no-go returned No-Go at 49.2%**, below the skill's 50% line, and every figure in it reproduced. The version 1.0 run (No-Go, 47.7%) had one wrong "what if" line.
- **The deal-breaker sat in the Q&A.** Q&A 175 made CJIS compliance (the FBI's security policy for criminal justice information) mandatory. The fictional company does not offer it, so the report applied the override: "The deal-breaker gives No-Go on its own, whatever the percentage."
- **The budget broke the bidder's own floor.** The State's $54,000 a year sat below the fictional company's $60,000 minimum, and the report recorded it as a break of that floor, since the tender sets no such gate.
- **One open item still needed a person.** The report scored a competitor question from Q&A clues while also listing it as open, although no file held competitor information.
Two habits follow for any tender: read the Q&A with the RFP, and keep only current files in `company/`.
## Should a Bid Manager Use Cowork or Claude Code?
Open Cowork when the job ends in a document you will edit and send, and open Claude Code when the tender is a folder you will come back to, rerun skills on and keep a written bid brief with.
Anthropic positions Cowork as Claude Code's way of working brought to knowledge work, with no terminal, and offers it in the desktop app on all paid plans (Help Center article [Get started with Claude Cowork](https://support.claude.com/en/articles/13345190-get-started-with-claude-cowork), November 2026). Our Cowork project instructions download is built for it, and Cowork is a fine place for most bid managers to start.
| | Cowork | Claude Code (desktop Code tab) |
| ------------------------ | ---------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------- |
| Where the work happens | An isolated cloud environment that reaches your local folders through the desktop app | On your computer, in the folder you open (Local environment) |
| What you set up | A task, with access to the folders it needs | A folder per tender, a CLAUDE.md bid brief and skills in .claude/skills |
| Permission modes | Manual, Auto and Skip (November 2026) | Auto, Manual, Accept edits and Plan (November 2026) |
| Skills and plugins | Skills enabled on your claude.ai account; plugins added here also work in chat and Claude Code | Project skills in the folder, personal skills on your computer, plugins, and in our test skills from your claude.ai account |
| Document formats | Not stated | Anthropic's pre-built Word, Excel, PowerPoint and PDF skills are not available |
| Plans | All paid plans (desktop app) | Pro, Max, Ultra, Team and Enterprise |
| Usage | Uses more of your limit than quick questions | Shares usage limits with Claude chat |
| Suits a bid manager when | The job ends in a document you edit and send | The tender is a folder you come back to and rerun skills on |
Four reasons a bid manager might pick the Code tab anyway:
- **The tender is already a folder.** Claude works on the files where they are.
- **The setup travels.** The bid brief and the project skills live in that folder.
- **You stay in control.** Manual and Plan modes let you approve each change or see the plan first.
- **Runs repeat.** The same skills and bid brief rerun on the next tender.
The trade-off: Anthropic's pre-built Word, Excel, PowerPoint and PDF skills are not available in Claude Code ([Agent Skills overview](https://platform.claude.com/docs/en/agents-and-tools/agent-skills/overview), November 2026). In our Code-tab test on November 10, 2026, Claude read a PDF directly but converted the Word files with a command first, and the converted text left out a spreadsheet embedded in the RFP's Word file. So ask Claude to pull out any embedded files and list them before it starts.
## Bring Your Approved Answers into Claude Code with AutoRFP.ai
AutoRFP.ai's MCP server lets Claude Code read a bid team's approved answers, each with its source, so the company side of a tender folder stops being a set of copies that go stale.
AutoRFP.ai drafts only from content your team has approved, shows the source behind every answer and routes anything it cannot support to a person. With AutoRFP.ai's MCP server connected, the bid brief can point Claude Code at that approved library instead of at copies.
The connection offers read scopes for looking up approved answers and their sources, and full scopes that let Claude Code write to projects and content in AutoRFP.ai. In both cases Claude Code acts as the signed-in user, and approvals stay inside AutoRFP.ai's own review workflow.
Setting up the connection is covered in [connecting Claude to the library that holds your approved answers](/blog/claude-connectors-for-rfps). For the platform behind it, see [AI proposal software built on approved answers](/ai-proposal-software).
## What to Do After Your First Tender Folder
Once one tender folder works, keep its bid brief and skills as the starting point for the next one, then add the next task.
- **Copy the skeleton.** Copy CLAUDE.md, `.claude/skills/` and the empty `tender/`, `company/` and `outputs/` folders for the next tender, then update the brief.
- **Add the next task.** On the same folder, [build a compliance matrix from the same tender](/blog/rfp-shredding-with-claude) with the Compliance Matrix Builder skill.
- **Make the scorecard yours.** [Write your own skill, such as a go/no-go scorecard](/blog/how-to-create-a-claude-skill-for-rfps), with your own weights and deal-breakers.
- **Plan for the team.** When the folder has to serve a whole bid team, with several writers, reviewers and approvals, read [how a Claude Code setup compares with purpose-built RFP software](/blog/chatgpt-or-claude-vs-rfp-software).