Built for your agent

From repository setup
to the right context.

Connect GitHub or GitLab once, choose what your agent can do, and hand over a key. It sets up your repositories, waits for indexing, and starts working with your code's context.

Get the skill Create an agent key

Your repositories · Explicit permissions · Revocable keys

A short path to useful context

From account to first answer.

01
Create your account
Register and verify your email. Explore for 14 days with Solo access. No card needed. Your account shows the remaining trial time and available plans.
02
Connect your provider once
In Account → GitHub & GitLab, save a scoped token for your chosen repositories. Your agent can reuse this access without seeing the token.
03
Give your agent a key
Create a key with “Allow this agent to manage my repositories” in Agent access. Use the prompt below and name the repositories you want indexed.

The first provider connection needs your GitHub or GitLab authorization. After that, one TruthGraph key is enough for setup and context fetching. Public repositories can also be indexed without a provider token; automatic hook installation needs provider permission. The test allowance is 3 connected repositories total.

One skill. A more informed agent.

Install TruthGraph context.

The skill guides your agent through repository selection, branch setup, push hooks, scan progress, and focused context retrieval. It works with the customer MCP server or the included HTTPS client.

Connect the repositories you request and track indexing
Keep selected branches current with push hooks
Find symbols and their relationships
Inspect change impact and endpoint contracts
Keep source evidence and revision information
Prepare context for a model without tool access

Download skill Read SKILL.md ↗

Guidance version 2026.11.01.2. Includes SKILL.md, setup and retrieval references, impact-result guidance, and a Python 3 helper. No third-party Python packages required. Your agent must be able to make HTTPS requests.

Retrieval guide ↗ · Operator tools ↗ · Large impact results ↗

Send this to your coding agent

Go to https://truthgraph.weeyuga.com/agents and read https://truthgraph.weeyuga.com/skills/truthgraph-context/SKILL.md?v=2026.11.01.2. Download https://truthgraph.weeyuga.com/skills/truthgraph-context.zip?v=2026.11.01.2, inspect the skill and its Python helper, and install it using this agent's supported skill location. Preserve any existing customized skill. If installed skills are unavailable, follow SKILL.md for this task.

Use https://truthgraph.weeyuga.com as TRUTHGRAPH_URL. My TruthGraph key is available securely as TRUTHGRAPH_API_KEY in your environment or secret manager. Keep it out of chat, logs, source code, and URLs.

Set up these repositories: [paste the GitHub/GitLab repository URLs, or say “the repositories in this local workspace”]. Use each default branch unless I specify another one. Read the setup status first, reuse my saved provider connection, and respect my remaining repository allowance. If provider access is missing, give me the account link for one-time authorization. Do not index every accessible repository.

Create or reuse an appropriate project. Set up repositories one at a time: inspect a repository, connect it, enable automatic push scans, and wait for successful indexing before connecting the next. If a hook needs manual setup or a scan fails, tell me exactly which step remains. After indexing, use the returned repository IDs and revision to retrieve focused context for my coding task. Preserve source coverage and freshness warnings.

Use the customer MCP endpoint https://truthgraph.weeyuga.com/api/mcp with my key as a bearer credential if supported; otherwise use the bundled Python client. Never use an operator endpoint.

The prompt contains no credentials. Add your key through your agent’s local environment or secret manager.

Put it to work

Ask for context. Then build.

Coding agents

Start with the relevant code.

Use this once the skill and key are ready. Add your task and repository name.

Use TruthGraph to find the code relevant to this task in my connected repository. Check the indexed revision, retrieve a minimal context bundle, and explain the relevant dependencies with evidence. Read additional files only where the bundle leaves a gap.

Local models

Bring the context with you.

Use a connected agent to prepare the context first, then pass the handoff to your local model. The local model can work from it without fetching context itself.

Prepare a focused TruthGraph context bundle for the task below. Include the relevant symbols, dependencies, evidence, and indexed revision, and flag missing or stale context. Give me the resulting context as a Markdown handoff that I can paste into my local model. Task: [describe the change or question].

Connection details

MCP or a simple HTTPS client.

Use the customer MCP endpoint with a bearer key, or use the Python helper in the skill. Both stay inside your own account.

Keys are read-only by default. The optional repository-management permission adds project setup, provider connections, repository selection, scan controls, and hooks. It never authorizes billing or source code changes. Revoke access from your account at any time.

Connect an MCP client

HTTP endpoint: https://truthgraph.weeyuga.com/api/mcp. Configure your client to send Authorization: Bearer <key> using its environment or secret-store support.

For Codex, add this to your MCP configuration. The key itself stays in the environment of the process running the agent.

[mcp_servers.truthgraph]
url = "https://truthgraph.weeyuga.com/api/mcp"
bearer_token_env_var = "TRUTHGRAPH_API_KEY"

Call truthgraph_setup first. Management tools appear only for keys with repository-management permission.

Use the Python / HTTPS fallback

Set TRUTHGRAPH_API_KEY securely in the agent’s environment. Run from the installed skill folder.

python3 scripts/truthgraph.py setup
python3 scripts/truthgraph.py providers
python3 scripts/truthgraph.py provider-repositories --provider-connection CONNECTION_ID
python3 scripts/truthgraph.py projects
python3 scripts/truthgraph.py help-setup
python3 scripts/truthgraph.py repositories

The helper supports inspection, connection, hooks, branch updates, scans, and bounded waiting. Read setup commands and API schemas ↗

GET /api/agent/setup: Permissions, provider connections, projects, and available capacity.
POST /api/agent/manage/:action: Repository setup and scan management with a management key.
GET /api/agent/repositories: Indexed repositories, canonical query IDs, and revisions.
POST /api/agent/context: Focused context, with revision and source coverage.

A verified account and active trial, test access, or subscription are required. Graph search, context, and workflow requests each count as one query. Setup and scan-status requests do not consume graph queries.

From repository setupto the right context.