by Coderica
Agents
Every agent. One process.
hashd does not replace your coding agents. It gives them a process. Claude Code, Codex and others are the workers; hashd is the foreman that scopes the work, checks it and keeps the record.
Supported agents
Bring the agents your team already uses.
Claude Code
Anthropic · the default for every stage
Codex
OpenAI
In preview:Gemini CLIGitHub CopilotOpenCodeKimi CodeQwen Code
Verified agents are tested end to end with hashd. Preview agents are configured and can be assigned with --force, but have not yet been verified end to end.
hashd agentsAgents------------------------------------------ claude [+] ~/.npm-global/bin/claude codex [+] ~/.npm-global/bin/codex copilot [x] not found gemini [+] ~/.npm-global/bin/gemini kimi [x] not found opencode [x] not found qwen [+] ~/.local/bin/qwen
Per stage
The right agent for each job.
Every stage declares the kind of invocation it needs, and any agent that supports it can fill the slot. Put one agent on implementation and another on review, per project or for the whole server.
Assignments are validated before they are saved: an agent that cannot serve a stage is refused, not discovered mid-run.
# implementation on Codex, review on Claudehashd project config set coder codexhashd project config set stage.review claude # every non-implement stage on one agenthashd project config set planner claude # one agent for every stage, server-widehashd config stages-use codex
On a team server
Everyone runs on their own account.
Agents run on the server, but the work belongs to people. A run acts as the workstream's owner, using a credential that person registered, stored encrypted at rest and never returned.
- Register oncehashd agents login claude reads the credential from your own machine where the agent has a supported way to, or asks you to paste it. Secrets are never passed as command-line flags.
- Your credential always winsA registered credential is used for your runs and overrides the server host's own agent login.
- No borrowingOn a team server, a run on your workstream refuses to start without your live credential, and says exactly which command fixes it.
- Dead credentials are rememberedWhen a run proves a credential expired or revoked, the next dispatch fails immediately with the recovery steps instead of burning another run.
| Agent | What you register |
|---|---|
| Claude | A claude setup-token token, or an API key |
| Codex | Your Codex login, or an API key |
| Gemini | Your Gemini login, or an API key |
| Copilot | Your GitHub CLI token, or a fine-grained token |
| Qwen | An API key and endpoint |
| Kimi | A Moonshot API key and model |
| OpenCode | A provider API key and model |
Steering
Your rules, in every prompt.
hashd decides when guidance reaches an agent, so agents see your rules exactly where they apply.
Directives
Operator guidance at three levels: global, project and workstream. Scope a block to specific stages such as review or implement.
Prompt overrides
Every prompt is a template you can list, read and override per project, then diff against the default or reset.
Grounded context
Agents query the Context Graph for modules, symbols, call sites and dependencies instead of rediscovering the codebase on every run.
Put your team's agents on the line.
One command installs hashd and everything it needs except git and a coding agent. Free in solo mode for individuals and teams of up to 10 engineers. Team mode comes with a commercial agreement. Pricing
curl -fsSL https://raw.githubusercontent.com/codr1/hashd-code/main/install.sh | bash
Rolling out to a team?
We work closely with a small group of design partners: hands-on onboarding, a direct line to the people building hashd, and a real say in what we build next.
Become a design partner